2  Extended Bibliography with Abstract Information

Abstracts are copyrighted by the respective owners and are used here as academic fair use.

2.1 Seneca Effect, Collapse, and Delayed Feedback Dynamics

Bardi (2011). The Limits to Growth Revisited. Springer. DOI: 10.1007/978-1-4419-9416-5

The Limits to Growth Revisited

AI Note: Bardi revisits the 1972 Limits to Growth (LTG) study four decades later, assessing the original model’s projections against empirical data on population, industrial output, pollution, and resource depletion through 2010. He argues that the LTG ‘standard run’ — which produces overshoot and collapse in the twenty-first century — tracks observed trends more closely than critics acknowledged, and that the underlying system-dynamics methodology remains a valid framework for understanding resource-constrained growth. For BDPD, this work provides the intellectual lineage connecting the project’s Seneca engine to the broader limits-to-growth tradition: the asymmetric collapse dynamics that BDPD³ operationalises on the Bardi ODE substrate are a formal descendant of the World3 model’s overshoot-and-decline structure. The book is primarily a retrospective validation exercise; it does not develop new models or governance prescriptions.

Bardi (2017). The Seneca Effect: Why Growth is Slow but Collapse is Rapid. Springer. DOI: 10.1007/978-3-319-57207-9

The Seneca Effect: Why Growth is Slow but Collapse is Rapid

The essence of this book can be found in a line written by the ancient Roman Stoic Philosopher Lucius Annaeus Seneca: “Fortune is of sluggish growth, but ruin is rapid”. This sentence summarizes the features of the phenomenon that we call “collapse,” which is typically sudden and often unexpected, like the proverbial “house of cards.” But why are such collapses so common, and what generates them? Several books have been published on the subject, including the well known “Collapse” by Jared Diamond (2005), “The collapse of complex societies” by Joseph Tainter (1998) and “The Tipping Point,” by Malcom Gladwell (2000). Why The Seneca Effect? This book is an ambitious attempt to pull these various strands together by describing collapse from a multi-disciplinary viewpoint. The reader will discover how collapse is a collective phenomenon that occurs in what we call today “complex systems,” with a special emphasis on system dynamics and the concept of “feedback.” From this foundation, Bardi applies the theory to real-world systems, from the mechanics of fracture and the collapse of large structures to financial collapses, famines and population collapses, the fall of entire civilzations, and the most dreadful collapse we can imagine: that of the planetary ecosystem generated by overexploitation and climate change. The final objective of the book is to describe a conclusion that the ancient stoic philosophers had already discovered long ago, but that modern system science has rediscovered today. If you want to avoid collapse you need to embrace change, not fight it. Neither a book about doom and gloom nor a cornucopianist’s dream, The Seneca Effect goes to the heart of the challenges that we are facing today, helping us to manage our future rather than be managed by it.

AI Note: Bardi’s book-length Report to the Club of Rome formalises the Seneca effect — the empirical regularity that complex systems grow slowly but collapse rapidly — through a three-variable ODE coupling resource (R), industrial capital (C), and pollution (P). Capital grows by consuming the resource; pollution is generated as a by-product of capital and degrades both; the characteristic behaviour is a Seneca cliff where capital rises slowly, peaks, and collapses much faster than it grew. This framework is the structural backbone of BDPD: the logistic baseline (paper_00) provides the symmetric starting point, while the Bardi ODE (paper_03, the Seneca engine) introduces true capital–pollution asymmetry, enabling the project’s central governance question — which signal triggers the regulator — to be posed and answered on a substrate where collapse speed is intrinsic, not shock-induced. The model treats collapse as a deterministic ODE outcome and abstracts from strategic agent interaction, the dimension BDPD’s LLM-agent experiments add.

Bardi (2018). A Seneca Collapse for the World’s Human Population?. Journal of Population and Sustainability, 2(2), 21–32. DOI: 10.3197/jps.2018.2.2.21

A Seneca Collapse for the World’s Human Population?

Most scenarios for the world’s human population predict continued growth into the 22nd century, while some indicate that it could stabilize or begin to fall before 2100. Almost always, decline is seen as not being faster than the preceding growth. Different scenarios are obtained if we consider the human population as a complex system, subject to the general rules that govern complex systems, in particular their tendency to show rapid changes which – in the case of populations – may take the shape of true collapses (defined here as “Seneca Collapses”). The present survey examines a small number of examples of rapid population collapses in the human and in the animal domains. While not pretending to be exhaustive, the data presented here show that biological populations do show rapid “Seneca-style” collapses. So, it is possible that the same phenomenon could occur for the world’s human population.

AI Note: Bardi extends the Seneca collapse concept to human population dynamics, surveying historical and natural examples of rapid population declines — from the St. Matthew Island reindeer to historical civilisations — that exhibited the characteristic slow-up, fast-down pattern. He argues that demographic projections that assume smooth stabilisation may miss the possibility of a Seneca-style population collapse driven by resource depletion and pollution feedback. BDPD³ and the mini-course cite this paper to establish the broader applicability of the Seneca framework beyond industrial-capital systems: the lesson that collapses are faster than growth, and that monitoring lagging signals is structurally insufficient, generalises from population dynamics to any system where resource consumption, growth, and delayed feedback interact. The paper is a survey of empirical cases; it does not provide a formal dynamical model or agent-based simulation.

Meadows et al. (1972). The Limits to Growth. Universe Books. ISBN: 0-87663-165-0

The Limits to Growth

Meadows et al. (2004). Limits to Growth: The 30-Year Update. Chelsea Green Publishing.

Limits to Growth: The 30-Year Update

AI Note: This third edition of the Limits to Growth updates the World3 model with thirty years of new data, concluding that humanity has already overshot Earth’s carrying capacity. The authors identify three necessary conditions for overshoot—growth, limits, and delays in perception and response—and present scenarios from collapse to a deliberate sustainability transition, warning that the window for corrective action narrows with each decade. BDPD’s Lesson 10 cites World3 alongside HANDY and the Seneca effect as a structural archetype: collapse is a recurrent pattern in systems where growth depletes a resource with delayed feedback. BDPD³’s leading-indicator experiment responds to this temporal challenge, testing whether a regulator armed on capital rather than pollution can avert collapse in a polycentric Seneca setting. A caveat: World3 operates at the global aggregate scale with no strategic agent behaviour. The micro-foundations of how heterogeneous agents respond to approaching limits—BDPD’s core experimental concern—lie outside its scope.

Motesharrei et al. (2014). Human and nature dynamics (HANDY): Modeling inequality and use of resources in the collapse or sustainability of societies. Ecological Economics, 101, 90–102. DOI: 10.1016/j.ecolecon.2014.02.014

Human and nature dynamics (HANDY): Modeling inequality and use of resources in the collapse or sustainability of societies

AI Note: Motesharrei, Rivas, and Kalnay present HANDY, a four-equation model coupling human population (Elites and Commoners), accumulated wealth, and natural resources within a predator–prey framework. The model shows that either ecological strain from resource over-depletion or economic stratification can independently trigger collapse, with two types: Type-L (Elite consumption starves Commoners) and Type-N (irreversible resource exhaustion). The HANDY framework directly informs BDPD³’s Bardi/Seneca ODE engine: the resource–capital–pollution coupling is the structural motif BDPD³ uses to test leading- versus lagging-indicator regulation. BDPD³ explicitly cites HANDY as the intellectual lineage from which the Seneca engine descends, noting the sign reversal (pollution degrades capital in Bardi, whereas reserves sustain population in HANDY) that produces the slow-growth, fast-collapse asymmetry. A caveat: HANDY was a qualitative thought experiment for stylised historical analogues. Its homogeneous production function abstracts away the strategic agent heterogeneity that is BDPD’s primary experimental variable.

Grammaticos et al. (2019). Revisiting the Human and Nature Dynamics model. arXiv preprint arXiv:1911.05533.

Revisiting the Human and Nature Dynamics model

AI Note: Grammaticos, Willox, and Satsuma present a simple ODE model for human-nature dynamics, formulated as a three-component system describing the interaction between a population, its resources, and reserves. The model exhibits rich dynamics — steady states, limit cycles, and collapse — with the important conclusion that collapse or near-collapse is indeed possible in such coupled systems. BDPD³’s model section cites this work as formally characterising the three-variable family on a HANDY-descended system, noting that the sign convention differs from the Bardi substrate: in Grammaticos et al., reserves sustain population, while in Bardi’s Seneca engine, pollution degrades capital. The paper’s finding that Seneca-style asymmetric collapse is achievable under some parameter choices motivates BDPD³’s structural OFAT robustness sweep. The model is analytical and does not incorporate strategic agents.

Perissi (2019). Highlighting the archetypes of sustainability management by means of simple dynamics models. Journal of Simulation. DOI: 10.1080/17477778.2019.1679612

Highlighting the archetypes of sustainability management by means of simple dynamics models

AI Note: Perissi presents a pedagogical suite of system-dynamics models—from Malthusian food–population dynamics through Lotka–Volterra predator–prey interactions to a simplified World Dynamics with a renewable energy extension—designed to teach sustainability concepts to non-specialists. The central claim is that these minimal archetypes capture the feedback structures governing resource overshoot, collapse, and the possibility of a steady-state transition through timely investment in sustainable infrastructure. BDPD³ cites this work as evidence that Bardi’s Seneca framework has been adopted for pedagogical modelling of resource exploitation, situating its own leading-indicator governance experiment within that lineage. Both share the premise that a signal’s temporal position relative to collapse determines whether governance can intervene. A caveat: Perissi’s models are qualitative illustrations, not parameterised predictions, and her homogeneous-world dynamics lack the strategic agent heterogeneity and polycentric structure that define BDPD’s experimental design.

Perissi & Bardi (2021). The Empty Sea: The Future of the Blue Economy. Springer. DOI: 10.1007/978-3-030-51898-1

The Empty Sea: The Future of the Blue Economy

AI Note: Perissi introduces the ‘empty sea’ metaphor — the gradual depletion of marine resources to the point of collapse — using system-dynamics modelling to illustrate how overexploitation proceeds invisibly until it is irreversible. The work builds on the Seneca framework, showing how resource extraction hides depletion until the system tips. BDPD⁰’s platform section cites Perissi alongside Bardi (2017) as part of the pedagogical lineage of encoding commons collapse dynamics into tangible forms — from system-dynamics models to card games. The Forest of Humbaba’s hidden reserve and Forest Die mechanisms operationalise the same invisible-depletion dynamic. The model is a system-dynamics demonstration; it does not incorporate strategic agents.

Dakos & Bascompte (2014). Critical slowing down as early warning for the onset of collapse in mutualistic communities. Proceedings of the National Academy of Sciences of the United States of America, 111. DOI: 10.1073/pnas.1406326111

Critical slowing down as early warning for the onset of collapse in mutualistic communities

Tipping points are crossed when small changes in external conditions cause abrupt unexpected responses in the current state of a system. In the case of ecological communities under stress, the risk of approaching a tipping point is unknown, but its stakes are high. Here, we test recently developed critical slowing-down indicators as early-warning signals for detecting the proximity to a potential tipping point in structurally complex ecological communities. We use the structure of 79 empirical mutualistic networks to simulate a scenario of gradual environmental change that leads to an abrupt first extinction event followed by a sequence of species losses until the point of complete community collapse. We find that critical slowing-down indicators derived from time series of biomasses measured at the species and community level signal the proximity to the onset of community collapse. In particular, we identify specialist species as likely the best-indicator species for monitoring the proximity of a community to collapse. In addition, trends in slowing-down indicators are strongly correlated to the timing of species extinctions. This correlation offers a promising way for mapping species resilience and ranking species risk to extinction in a given community. Our findings pave the road for combining theory on tipping points with patterns of network structure that might prove useful for the management of a broad class of ecological networks under global environmental change.

Sterman (1989). Misperceptions of feedback in dynamic decision making. Organizational Behavior and Human Decision Processes, 43(3), 301–335. DOI: 10.1016/0749-5978(89)90041-1

Misperceptions of feedback in dynamic decision making

In recent years laboratory experiments have shed significant light on the behavior of economic agents in a variety of microeconomic and decision-theoretic contexts such as auction markets, portfolio choice, and preference elicitation. Despite the success of experimental techniques in the micro domain, there has been relatively little work linking the behavior of decision makers to the dynamics of larger organizations such as corporations, industries, or the macroeconomy. This paper presents a laboratory experiment in which subjects manage a simulated economy. Subjects must invest sufficient capital plant and equipment to satisfy demand. Subjects were given complete and perfect information regarding the structure of the simulated economy, the values of all variables, and the past history of the system. Nevertheless, the overwhelming majority of the subjects generate significant and costly oscillations. A simple decision rule based on the anchoring and adjustment heuristic is shown to simulate the subjects’ decisions quite well. Several distinct sources of the subjects’ poor performance are identified and termed “misperceptions of feedback.” The decision rule is related to various models of economic fluctuations; implications for experimental investigation of dynamic decision making in aggregate systems are explored.

Sterman (2000). Business Dynamics: Systems Thinking and Modeling for a Complex World. McGraw-Hill. ISBN: 9780072311358

Business Dynamics: Systems Thinking and Modeling for a Complex World

This comprehensive textbook introduces systems thinking and system dynamics as rigorous frameworks for analyzing complex business, economic, and social systems. Sterman covers feedback loops, stocks and flows, time delays, nonlinearities, and policy design, integrating theory with hands-on modeling exercises and real-world case studies. The book equips readers with tools to diagnose dynamic complexity, avoid common decision traps, and design robust strategies for sustainable performance in rapidly changing environments.

Moxnes (2000). Not only the tragedy of the commons: misperceptions of bioeconomics. System Dynamics Review, 16(4), 325–340. DOI: 10.1287/mnsc.44.9.1234

Not only the tragedy of the commons: misperceptions of bioeconomics

An exploratory search for explanations of mismanagement of renewable resources, other than the theory of the commons, was performed by an experiment. Eighty three subjects, mostly recruited from the fisheries sector in Norway, were asked to manage the same simulated virgin fish stock, one subject at a time. Exclusive property rights were granted to rule out the commons problem. Despite perfect property rights, subjects consistently overinvested, leading to an average overcapacity of 60%. The resource was reduced by an average of 15% below its optimal level. Overcapacity and tough “quotas” resemble the situation in Norwegian and other fisheries during the past few decades. The likely explanation of the observed behaviour is misperception of feedback, a phenomenon that occurs in many experimental studies of dynamically complex systems. Such misperceptions add a new and important dimension to the problem of renewable resource management, beyond the commons problem.

Santos & Pacheco (2011). Risk of collective failure provides an escape from the tragedy of the commons. Proceedings of the National Academy of Sciences, 108(26), 10421–10425. DOI: 10.1073/pnas.1015648108

Risk of collective failure provides an escape from the tragedy of the commons

From group hunting to global warming, how to deal with collective action may be formulated in terms of a public goods game of cooperation. In most cases, contributions depend on the risk of future losses. Here, we introduce an evolutionary dynamics approach to a broad class of cooperation problems in which attempting to minimize future losses turns the risk of failure into a central issue in individual decisions. We find that decisions within small groups under high risk and stringent requirements to success significantly raise the chances of coordinating actions and escaping the tragedy of the commons. We also offer insights on the scale at which public goods problems of cooperation are best solved. Instead of large-scale endeavors involving most of the population, which as we argue, may be counterproductive to achieve cooperation, the joint combination of local agreements within groups that are small compared with the population at risk is prone to significantly raise the probability of success. In addition, our model predicts that, if one takes into consideration that groups of different sizes are interwoven in complex networks of contacts, the chances for global coordination in an overall cooperating state are further enhanced.

AI Note: Santos and Pacheco use evolutionary game theory to model an N-person threshold dilemma where risk of collective failure makes cooperation viable when it would otherwise be dominated by defection. Small groups under stringent requirements outperform large ones; stochastic effects in finite populations allow tunnelling through coordination barriers, making cooperation the prevalent strategy. This work underpins BDPD⁰’s finding of a discontinuous collapse threshold (r ≤ 0.32): a single aggressive agent suffices, because the Arena’s Gate+Rank victory embeds the threshold uncertainty Santos and Pacheco identify as making collective action fragile. BDPD⁰ further operationalises this through its Hidden Reserve and Forest Die mechanism. A caveat: the Santos–Pacheco model treats cooperation as binary in a one-shot game with fixed group composition, whereas BDPD involves iterated extraction with wealth accumulation. The model’s group-size and network-topology insights do not directly translate to settings where extraction intensity varies.

2.2 Tragedy of the Commons and Institutional Governance

Hardin (1968). The Tragedy of the Commons. Science, 162(3859), 1243–1248. DOI: 10.1126/science.162.3859.1243

The Tragedy of the Commons

The population problem has no technical solution; it requires a fundamental extension in morality.

AI Note: Hardin argues that problems such as population growth and environmental degradation have no technical solution and require changes in values and institutions. Through the parable of a pasture open to all, he shows that rational herders each add cattle because private benefits are fully captured while overgrazing costs are shared, dooming the commons. He extends the logic to pollution and population, concluding that “mutual coercion, mutually agreed upon” is the necessary remedy. BDPD⁰ adopts this framing: the Arena’s Gate+Rank victory maps onto Hardin’s tension between private gain and collective preservation, and the step-collapse threshold (P1–P5) gives the parable empirical precision under realistic parameters. BDPD² extends the logic to nested arenas where externalities cascade, while Lesson 3 of the mini-course traces the Hardin-to-Ostrom arc explicitly. A caveat: the essay assumes homogeneous actors with costless access and no regeneration dynamics—the exact simplifications BDPD relaxes by making heterogeneity, resource dynamics, and information structure first-class variables.

Ostrom (1990). Governing the Commons: The Evolution of Institutions for Collective Action. Cambridge University Press. DOI: 10.1017/CBO9780511807763

Governing the Commons: The Evolution of Institutions for Collective Action

The governance of natural resources used by many individuals in common is an issue of increasing concern to policy analysts. Both state control and privatisation of resources have been advocated, but neither the state nor the market have been uniformly successful in solving common pool resource problems. Offering a critique of the foundations of policy analysis as applied to natural resources, Elinor Ostrom here provides a unique body of empirical data to explore conditions under which common pool resource problems have been satisfactorily or unsatisfactorily solved. Dr Ostrom first describes three models most frequently used as the foundation for recommending state or market solutions. She then outlines theoretical and empirical alternatives to these models in order to illustrate the diversity of possible solutions. In the following chapters she uses institutional analysis to examine different ways - both successful and unsuccessful - of governing the commons. In contrast to the proposition of the tragedy of the commons argument, common pool problems sometimes are solved by voluntary organisations rather than by a coercive state. Among the cases considered are communal tenure in meadows and forests, irrigation communities and other water rights, and fisheries.

AI Note: Through comparative analysis of long-enduring CPR institutions—Swiss alpine meadows, Japanese forests, Spanish huertas—Ostrom demonstrates that communities can self-govern shared resources without state or market control. She distils eight design principles, from clearly defined boundaries and graduated sanctions to nested enterprises, and critiques the metaphorical misuse of Hardin, the prisoner’s dilemma, and Olson as policy foundations. BDPD¹ operationalises principles 3–5 as programmable governance surfaces; its D2/D3 pilots confirm that graduated sanctions preserve the commons where flat sanctions fail, because escalation shape rather than amount is the operative mechanism. BDPD² extends the nested-enterprise principle to multi-arena worlds, while BDPD³ maps monitoring onto the leading-indicator governance question. A caveat: Ostrom’s cases involved human communities with shared cultural norms and generational horizons. Whether the design principles retain their force when appropriators are heuristic or LLM agents on abstract logistic substrates is the question BDPD tests rather than assumes.

Ostrom (2005). Understanding Institutional Diversity. Princeton University Press.

Understanding Institutional Diversity

The analysis of how institutions are formed, how they operate and change, and how they influence behavior in society has become a major subject of inquiry in politics, sociology, and economics. A leader in applying game theory to the understanding of institutional analysis, Elinor Ostrom provides in this book a coherent method for undertaking the analysis of diverse economic, political, and social institutions. Understanding Institutional Diversity explains the Institutional Analysis and Development (IAD) framework, which enables a scholar to choose the most relevant level of interaction for a particular question. This framework examines the arena within which interactions occur, the rules employed by participants to order relationships, the attributes of a biophysical world that structures and is structured by interactions, and the attributes of a community in which a particular arena is placed. The book explains and illustrates how to use the IAD in the context of both field and experimental studies. Concentrating primarily on the rules aspect of the IAD framework, it provides empirical evidence about the diversity of rules, the calculation process used by participants in changing rules, and the design principles that characterize robust, self-organized resource governance institutions.

Ostrom (2009). A general framework for analyzing sustainability of social-ecological systems. Science, 325(5939), 419–422. DOI: 10.1126/science.1172133

A general framework for analyzing sustainability of social-ecological systems

A major problem worldwide is the potential loss of fisheries, forests, and water resources. Understanding of the processes that lead to improvements in or deterioration of natural resources is limited, because scientific disciplines use different concepts and languages to describe and explain complex social-ecological systems (SESs). Without a common framework to organize findings, isolated knowledge does not cumulate. Until recently, accepted theory has assumed that resource users will never self-organize to maintain their resources and that governments must impose solutions. Research in multiple disciplines, however, has found that some government policies accelerate resource destruction, whereas some resource users have invested their time and energy to achieve sustainability. A general framework is used to identify 10 subsystem variables that affect the likelihood of self-organization in efforts to achieve a sustainable SES.

Ostrom et al. (1994). Rules, Games, and Common-Pool Resources. University of Michigan Press. ISBN: 978-0-472-06561-5

Rules, Games, and Common-Pool Resources

Explores ways that the tragedy of the commons can be avoided by people who use common-property resources

Dietz et al. (2003). The struggle to govern the commons. Science, 302(5652), 1907–1912. DOI: 10.1126/science.1091015

The struggle to govern the commons

Human institutions–ways of organizing activities–affect the resilience of the environment. Locally evolved institutional arrangements governed by stable communities and buffered from outside forces have sustained resources successfully for centuries, although they often fail when rapid change occurs. Ideal conditions for governance are increasingly rare. Critical problems, such as transboundary pollution, tropical deforestation, and climate change, are at larger scales and involve nonlocal influences. Promising strategies for addressing these problems include dialogue among interested parties, officials, and scientists; complex, redundant, and layered institutions; a mix of institutional types; and designs that facilitate experimentation, learning, and change.

Poteete et al. (2010). Working Together: Collective Action, the Commons, and Multiple Methods in Practice. Princeton University Press. DOI: 10.1515/9781400835157

Working Together: Collective Action, the Commons, and Multiple Methods in Practice

Working Together examines how different methods have promoted various theoretical developments related to collective action and the commons, and demonstrates the importance of cross-fertilization involving multimethod research across traditional boundaries. The authors look at why cross-fertilization is difficult to achieve, and they show ways to overcome these challenges through collaboration. The authors provide numerous examples of collaborative, multimethod research related to collective action and the commons. They examine the pros and cons of case studies, meta-analyses, large-N field research, experiments and modeling, and empirically grounded agent-based models, and they consider how these methods contribute to research on collective action for the management of natural resources. Using their findings, the authors outline a revised theory of collective action that includes three elements: individual decision making, microsituational conditions, and features of the broader social-ecological context.

Cox et al. (2010). A Review of Design Principles for Community-Based Natural Resource Management. Ecology and Society, 15(4), 38. DOI: 10.5751/ES-03704-150438

A Review of Design Principles for Community-Based Natural Resource Management

In 1990, Elinor Ostrom proposed eight design principles, positing them to characterize robust institutions for managing common-pool resources such as forests or fisheries. Since then, many studies have explicitly or implicitly evaluated these design principles. We analyzed 91 such studies to evaluate the principles empirically and to consider what theoretical issues have arisen since their introduction. We found that the principles are well supported empirically and that several important theoretical issues warrant discussion. We provide a reformulation of the design principles, drawing from commonalities found in the studies.

Malézieux & Spiegelman (2025). An anatomical review of the common pool resource game. Experimental Economics, 28(3), 468-–491. DOI: 10.1017/eec.2024.6

An anatomical review of the common pool resource game

Over the last four decades, a broad stream of experimental literature has been published using the Common Pool Resource (CPR) game to study how people react to congestible resources, and how to keep such resources from socially harmful overexploitation. With the goal of providing guidance to future work on this still-important paradigm, we provide a narrative review of the literature, summarizing the results for several key aspects of the experimental operationalization. We classify these aspects into two broad categories. The first describes ‘environmental’ assumptions on the modeled resource problem itself. This refers to aspects of the experimental environment reflecting factors such as group size, resource size and asymmetry of access, which generally constitute the nature of the dilemma. The second category involves ‘institutional’ issues related to how people might solve the problem, such as user communication between subjects, information about previous subjects’ choices, and regulatory measures.

AI Note: Malézieux and Spiegelman review 123 experimental CPR studies, classifying design components into environmental assumptions (group size, resource size, regeneration rate, uncertainty) and institutional issues (communication, feedback, quotas, punishments). The headline: communication is the most robustly powerful single measure for CPR management, with all 24 studies testing it finding positive effects on sustainability and payoffs. Endogenous peer punishment shows little systematic impact. The BDPD mini-course’s Lesson 1 uses this review to anchor the claim that cheap talk has a large cooperation effect among humans, then contrasts it with BDPD¹’s D1 result: when LLM agents replace human subjects, the cheap-talk effect collapses to a negligible Cohen’s d of 0.17. Lesson 2 draws on the group-size findings to qualify Olson’s claims. A caveat: the review covers human subjects almost exclusively from WEIRD populations. Its conclusions about communication are bounded to that population; extending them to computational agents is the extrapolation BDPD tests.

Tisserand et al. (2022). Management of common pool resources in a nation-wide experiment. Ecological Economics, 201, 107566. DOI: 10.1016/j.ecolecon.2022.107566

Management of common pool resources in a nation-wide experiment

Dilemmas related to the use of environmental resources concern diverse populations at local or global scales. Frequently, individuals are unable to visualize the consequences of their actions, where they belong in the decision-making line, and have no information about past choices or the time horizon. We design a new one-shot extraction game to capture these dynamic decisions. We present results from a nationwide common pool resource experiment, conducted simultaneously in eleven French cities, involving a total of 2813 participants. We examine, for the first time, the simultaneous impact of several variables on the amount of resource extracted: the local vs. the national scale of the resource, the size of the group (small vs. big), the low vs. high recovery rate of the resource, and the available information. We show that individuals significantly reduce extraction levels in local as compared to national level dilemmas and that providing recommendations on sustainable extraction amounts significantly improves the sustainability of the resource. Overall, women extract significantly less, but care more about preserving the local resource; older participants extract significantly more resources but extract less from the national resource. Our experiment also fulfills a science popularization pedagogical aim, which we discuss.

Cárdenas et al. (2015). Stable Sampling Equilibrium in Common Pool Resource Games. Games, 6(3), 299–318. DOI: 10.3390/g6030299

Stable Sampling Equilibrium in Common Pool Resource Games

This paper reconsiders evidence from experimental common pool resource games from the perspective of a model of payoff sampling. Despite being parameter-free, the model is able to replicate some striking features of the data, including single-peaked frequency distributions, the persistent use of strictly dominated actionsand stable heterogeneity in choices. These properties can also be accurately replicated using logit quantal response equilibrium (QRE), but only by tuning the free parameter separately for separate games. When the QRE parameter is constrained to be the same across games, sampling equilibrium provides a superior fit to the data. We argue that these findings are likely to generalize to other complex games with multiple players and strategies.

Janssen & Ostrom (2006). Empirically Based, Agent-based Models. Ecology and Society, 11(2). URL: http://www.ecologyandsociety.org/vol11/iss2/art37/

Empirically Based, Agent-based Models

There is an increasing drive to combine agent-based models with empirical methods. An overview is provided of the various empirical methods that are used for different kinds of questions. Four categories of empirical approaches are identified in which agent-based models have been empirically tested: case studies, stylized facts, role-playing games, and laboratory experiments. We discuss how these different types of empirical studies can be combined. The various ways empirical techniques are used illustrate the main challenges of contemporary social sciences: (1) how to develop models that are generalizable and still applicable in specific cases, and (2) how to scale up the processes of interactions of a few agents to interactions among many agents.

2.3 Wealth Distribution and Inequality

Piketty (2014). Capital in the Twenty-First Century. Harvard University Press. DOI: 10.4159/9780674369542

Capital in the Twenty-First Century

In this landmark work, Piketty analyzes data from twenty countries over two centuries to reveal key economic and social patterns. He argues that when the rate of return on capital (r) exceeds the rate of economic growth (g), wealth tends to concentrate at the top, leading to unsustainable levels of inequality. The book proposes a global progressive tax on wealth as a policy solution to mitigate these trends and preserve democratic values.

AI Note: Piketty’s landmark book documents the historical tendency of wealth to concentrate when the rate of return on capital exceeds the rate of economic growth (r > g), drawing on centuries of tax data across multiple countries. For BDPD, this provides the macro-historical backdrop to its micro-foundational finding that wealth-capacity feedback — where harvest increases with accumulated wealth (1 + 0.05 × w) — compounds inequality within a single commons. The BDPD platform’s Gini tracking and welfare-score metric operationalise Piketty’s distributional lens at the agent level. The book addresses national-level wealth concentration over centuries; it does not model the strategic dynamics of common-pool resource extraction.

Wilkinson & Pickett (2009). The Spirit Level: Why More Equal Societies Almost Always Do Better. Allen Lane. ISBN: 978-1-84614-039-6

The Spirit Level: Why More Equal Societies Almost Always Do Better

This book presents evidence that income inequality has profound negative effects on health and social well-being. By comparing data from developed nations and US states, the authors show that more equal societies have better outcomes in life expectancy, mental health, obesity, crime rates, and social mobility. They argue that inequality erodes social cohesion and trust, creating a `social gradient’ of disadvantage that affects everyone, not just the poor.

Boyce (1994). Inequality as a Cause of Environmental Degradation. Ecological Economics, 11(1), 43–55. DOI: 10.1016/0921-8009(94)90198-8

Inequality as a Cause of Environmental Degradation

This paper advances two hypotheses. First, the extent of an environmentally degrading economic activity is a function of the balance of power between the winners, who derive net benefits from the activity, and the losers, who bear net costs. Second, greater inequalities of power and wealth lead, all else equal, to more environmental degradation.

Chakraborti & Chakrabarti (2000). Statistical Mechanics of Money: How Saving Propensity Affects Its Distribution. European Physical Journal B, 17, 167–170. DOI: 10.1007/s100510070173

Statistical Mechanics of Money: How Saving Propensity Affects Its Distribution

We consider a simple model of a closed economic system where the total money is conserved and the number of economic agents is fixed. Analogous to statistical systems in equilibrium, money and the average money per economic agent are equivalent to energy and temperature, respectively. We investigate the effect of the saving propensity of the agents on the stationary or equilibrium probability distribution of money. When the agents do not save, the equilibrium money distribution becomes the usual Gibb’s distribution, characteristic of non-interacting agents. However with saving, even for individual self-interest, the dynamics becomes cooperative and the resulting asymmetric Gaussian-like stationary distribution acquires global ordering properties. Intriguing singularities are observed in the stationary money distribution in the market, as functions of the marginal saving propensity of the agents.

Chakraborti et al. (2011). Econophysics review: I. Empirical facts. Quantitative Finance, 11(7), 991–1012. DOI: 10.1080/14697688.2010.539248

Econophysics review: I. Empirical facts

This article and the companion paper aim at reviewing recent empirical and theoretical developments usually grouped under the term Econophysics. Since the name was coined in 1995 by merging the words ‘Economics’ and ‘Physics’, this new interdisciplinary field has grown in various directions: theoretical macroeconomics (wealth distribution), microstructure of financial markets (order book modeling), econometrics of financial bubbles and crashes, etc. We discuss the interactions between Physics, Mathematics, Economics and Finance that led to the emergence of Econophysics. We then present empirical studies revealing the statistical properties of financial time series. We begin the presentation with the widely acknowledged ‘stylized facts’, which describe the returns of financial assets—fat tails, volatility clustering, autocorrelation, etc.—and recall that some of these properties are directly linked to the way ‘time’ is taken into account. We continue with the statistical properties observed on order books in financial markets. For the sake of illustrating this review, (nearly) all the stated facts are reproduced using our own high-frequency financial database. Finally, contributions to the study of correlations of assets such as random matrix theory and graph theory are presented. The companion paper will review models in Econophysics from the point of view of agent-based modeling.

Ciešła & Snarska (2020). A Simple Mechanism Causing Wealth Concentration. Entropy, 22(10), 1148. DOI: 10.3390/e22101148

A Simple Mechanism Causing Wealth Concentration

We study mechanisms leading to wealth condensation. As a natural starting point, our model adopts a neoclassical point of view, i.e., we completely ignore work, production, and productive relations, and focus only on bilateral link between two randomly selected agents. We propose a simple matching process with deterministic trading rules and random selection of trading agents. Furthermore, we also neglect the internal characteristic of traded goods and analyse only the relative wealth changes of each agent. This is often the case in financial markets, where a traded good is money itself in various forms and various maturities. We assume that agents trade according to the rules of utility and decision theories. Agents possess incomplete knowledge about market conditions, but the market is in equilibrium. We show that these relatively frugal assumptions naturally lead to a wealth condensation. Moreover, we discuss the role of wealth redistribution in such a model.

La Porta & Zapperi (2025). Persistence of wealth inequality from network effects. PLOS Complex Systems, 2(6), e0000057. DOI: 10.1371/journal.pcsy.0000050

Persistence of wealth inequality from network effects

Addressing wealth and income inequality requires a thorough understanding of the mechanisms driving these disparities. Agent-Based Models (ABMs) offer a powerful tool for simulating these complex systems, capturing the intricate interplay of individual behaviors and emergent macroeconomic trends. Here we consider two existing ABM classes: one, exemplified by the Nirei-Souma (NS) model, which simulates how individuals accumulate wealth through income from work, returns on investments, and consumption, and the other, represented by the Bouchaud-Mezard (BM) model, which emphasizes the role of wealth exchanges and random returns in shaping the wealth distribution. Drawing on empirical evidence of wealth and income distribution in Italy, we benchmark both models revealing that they effectively captures Pareto-like wealth distribution, but fail to fully account for the persistent lack of social mobility observed in empirical data. To overcome this limitation, we propose an interacting version of the NS model, integrating it with wealth exchange mechanisms. Through this interacting model, we can show the influence of network topology on wealth distribution and dynamics. Simulations on hierarchical networks yield results that align more closely with empirical observations compared to regular random graphs, highlighting the importance of hierarchical interactions in shaping wealth inequality and social mobility. The model is further analyzed to reveal the interplay between income sources and wealth accumulation.

Villafañe et al. (2025). Wealth inequality in agent-based economies: The dominant role of social protection over growth. Physica A, 661, 130428. DOI: 10.1016/j.physa.2025.131053

Wealth inequality in agent-based economies: The dominant role of social protection over growth

Persistent wealth inequality, where a small fraction of the population accumulates most resources while the majority remains economically vulnerable, is a widespread phenomenon. We investigate its underlying mechanisms using an agent-based Yard-Sale model that incorporates two complementary features: transaction rules that favor poorer agents, representing social protection policies, and an economic growth process with explicit wealth redistribution. Our results reveal that social protection plays a dominant role in reducing inequality, while redistribution primarily serves to reintegrate excluded agents. These findings suggest that social protection policies, that is, targeted mechanisms favoring economically vulnerable agents, may have a substantially greater impact on reducing inequality than redistribution driven solely by economic growth. We also find that both the shape of the wealth distributions and the resulting inequality levels are strongly influenced by the underlying distribution of individual risk, highlighting the importance of considering agent heterogeneity when modeling economic dynamics.

Liu (2025). How Fixed-Amount Transactions and Liquidity Constraints Amplify Wealth Inequality. URL: https://arxiv.org/html/2511.08202

How Fixed-Amount Transactions and Liquidity Constraints Amplify Wealth Inequality

This paper investigates the emergence of wealth inequality through a minimalist kinetic exchange model that incorporates two fundamental economic features: fixed-amount transactions and hard budget constraints. In contrast to the maximum entropy principle, which predicts an exponential Boltzmann-Gibbs distribution with moderate inequality (Gini ≈ 0.333) for unconstrained wealth exchange, we demonstrate that these realistic trading rules drive the system toward a highly unequal steady state. We develop a self-consistent mean-field theory, deriving a master equation where agent income follows a Poisson process coupled to the poverty rate. Numerical solution reveals a stationary distribution characterized by a substantial pauper class (p0≈55.1%), high Gini coefficient (G≈0.649), and exponential tail—significantly deviating from the maximum entropy benchmark. Agent-based simulations confirm these findings (p0≈45.2%, G≈0.618). We identify the poverty trap as the key mechanism: the liquidity constraint creates asymmetric economic agency, where zero-wealth agents become passive recipients, unable to participate in wealth circulation. This work establishes that substantial inequality can emerge spontaneously from equal-opportunity exchanges under basic economic constraints, without requiring agent heterogeneity or multiplicative advantage, providing a mechanistic foundation for understanding poverty as an emergent property of exchange rules.

Bagatella-Flores et al. (2015). Wealth distribution of simple exchange models coupled with extremal dynamics. Physica A: Statistical Mechanics and its Applications, 417, 168–175. DOI: 10.1016/j.physa.2014.07.081

Wealth distribution of simple exchange models coupled with extremal dynamics

Punctuated Equilibrium (PE) states that after long periods of evolutionary quiescence, species evolution can take place in short time intervals, where sudden differentiation makes new species emerge and some species extinct. In this paper, we introduce and study the effect of punctuated equilibrium on two different asset exchange models: The yard sale model (YS, winner gets a random fraction of a poorer player’s wealth) and the theft and fraud model (TF, winner gets a random fraction of the loser’s wealth). The resulting wealth distribution is characterized using the Gini index. In order to do this, we consider PE as a perturbation with probability of being applied. We compare the resulting values of the Gini index at different increasing values of in both models. We found that in the case of the TF model, the Gini index reduces as the perturbation increases, not showing dependence with the agents number. While for YS we observe a phase transition which happens around . For perturbations the Gini index reaches the value of one as time increases (an extreme wealth condensation state), whereas for perturbations bigger or equal than the Gini index becomes different to one, avoiding the system reaches this extreme state. We show that both simple exchange models coupled with PE dynamics give more realistic results. In particular for YS, we observe a power low decay of wealth distribution.

Vallejos & others (2017). An Agent-Based Model of Wealth Distribution in the US. Journal of Economic Interaction and Coordination, 12(3), 567–590. DOI: 10.1007/s11403-017-0200-9

An Agent-Based Model of Wealth Distribution in the US

Pareto cautiously asserted that the wealth and income distributions which bear his name are universal, basing his argument on observations of this distribution in many different types of economies. In this paper, we present an agent based model (and a scalable approximation of it) in a closely related spirit. The central feature of this model is that wealth enables an individual to secure more wealth. Specifically, the important and novel feature of this model is its ability to simultaneously produce both the Pareto distribution observed in empirical data for the top 10% of the population and the exponential distribution observed for the lower 90%. We show that the model produces these distributions of wealth when initialized with an equitable distribution. Then, using historical data, we initialize the model with US wealth shares in 1988 and show that the model tracks wealth share changes from 1988 to 2012. Simulations to 2088 project that the top 0.01% of the population will possess more than 70% of the total wealth in the economy.

Dayton-Johnson & Bardhan (2002). Inequality and Conservation on the Local Commons: A Theoretical Exercise. The Economic Journal, 112(481), 577–602. DOI: 10.1111/1468-0297.00731

Inequality and Conservation on the Local Commons: A Theoretical Exercise

To analyse the effect of asset inequality on co‐operation within a group, we consider a two‐player nonco‐operative model of conservation of a common‐pool resource. Overexploitation by one user affects another’s payoff by reducing the next‐period catch. We give necessary and sufficient conditions such that conservation is a Nash equilibrium, and show that increasing inequality does not, in general, favour full conservation. However, once inequality is sufficiently great, further inequality can raise efficiency. Thus, the relationship between inequality and economic efficiency is U‐shaped. Finally, we analyse the implications for conservation if players have earning opportunities outside the commons.

AI Note: Dayton-Johnson and Bardhan analyse the effect of asset inequality on cooperation in a two-player noncooperative model of common-pool resource conservation. They find a U-shaped relationship: at moderate inequality, coordination fails and conservation is not an equilibrium, but at sufficiently high inequality, the wealthier player has enough individual stake to conserve unilaterally — the ‘Olson effect.’ BDPD⁰’s introduction cites this model to anchor the proposition that inequality does not monotonically degrade cooperation: the project’s wealth-capacity feedback (harvest scales as 1 + 0.05 × w) operationalises the compounding advantage the model describes. The BDPD platform then tests whether this U-shape survives in the richer strategic environment of multi-agent, multi-round extraction under a logistic or Seneca resource. The model is a two-player analytical exercise with a fixed payoff structure; it does not model the dynamics of cumulative extraction or the stochastic collapse thresholds central to BDPD.

Yoon & Armsworth (2025). The Role of Wealth Inequality in Community Management of a Common-Pool Resources Through Voluntary Sanctioning. Natural Resource Modeling, 38(1), e70001. DOI: 10.1111/nrm.70001

The Role of Wealth Inequality in Community Management of a Common-Pool Resources Through Voluntary Sanctioning

Community-based regulation of common-pool resources like fish stocks, irrigation, and clean air, depends on establishing rules to prevent over-exploitation and sanctioning free-riders. However, persistence of these social arrangements can be challenged when some individuals enjoy greater access to the common-pool resource because of differences in wealth. Here, we investigated how wealth inequality impacts the management of common-pool resources in self-regulating communities that sanction over-exploiting free-riders. We used a game theoretic model to isolate the effect of inequality on resource management. Specifically, we employed a model where players decide both how much to contribute to the resource and whether to sanction those who contribute less than the socially agreed-upon amount. Through Nash equilibrium analysis, our results showed a U-shaped relationship between wealth inequality and resource sustainability: high sustainable provision of resources was achieved under conditions of both high inequality and high equality. Resources were cooperatively maintained under equality via sanctioning, but maintained solely by the wealthiest individual under high inequality through privatization of resources. Increased severity of imposed sanctions expanded the range of wealth inequality in which resources were managed cooperatively through the voluntary sanctioning mechanism. Conversely, the increased cost of imposing sanctions decreased the range of wealth inequality where voluntary sanctioning was employed. The impact of wealth inequality on community-management of common-pool resources appears to defy simple generalizations about there being trade-offs between equity and efficiency of resource use.

AI Note: Yoon demonstrates that voluntary sanctioning mechanisms can restore the sustainability of the commons, expanding the range of inequality levels under which resources can be managed cooperatively. BDPD⁰’s introduction and discussion cite this work as evidence that institutional remedies — specifically, Ostrom-style graduated sanctions — can counteract the inequality-driven degradation of commons. BDPD¹’s graduated-sanction pilots (D2, D3) directly operationalise Yoon’s finding: the ladder-sweep confirms that escalation shape (graduated), not constant amount (flat), drives the sanctioning effect, extending Yoon’s voluntaristic framework to an LLM-agent substrate with institutional (pact-based) rather than peer-to-peer sanctioning. The study uses human-subject experiments; it does not test sanctioning with LLM agents or on a Seneca substrate.

2.4 Differential Games and Evolutionary Game Theory

Dockner & others (2000). Differential Games in Economics and Management Science. Cambridge University Press. DOI: 10.1017/CBO9780511805127

Differential Games in Economics and Management Science

This comprehensive textbook provides a rigorous introduction to differential games, focusing on applications in economics and management science. It covers both open-loop and feedback Nash equilibria, dynamic programming techniques, and stochastic differential games. The book serves as a foundational reference for researchers and graduate students interested in modeling strategic interactions over time in continuous settings.

Başar & Olsder (1999). Dynamic Noncooperative Game Theory. SIAM. DOI: 10.1137/1.9781611971132

Dynamic Noncooperative Game Theory

A classic text in the field, this book offers a thorough treatment of dynamic noncooperative game theory. It systematically develops the theory of zero-sum and nonzero-sum differential games, including linear-quadratic structures, singular perturbations, and hierarchical games. The second edition includes updated material on robust control and H-infinity optimization, making it essential reading for engineers and economists alike.

Sethi (2021). Differential Games. Springer. DOI: 10.1007/978-3-030-91745-6_13

Differential Games

When there are more than a single decision-maker, each having one’s own objective function that each is trying to maximize, subject to a set of differential equations, then we require an extension of the optimal control theory referred to as the theory of differential games. While representing a generalization of optimal control problems in cases where there is more than one controller or player, differential games are conceptually far more complex than optimal control problems in the sense that it is no longer obvious what constitutes a solution. Indeed, there are different types of solutions such as minimax, Nash, and Stackelberg. We discuss minimax solutions for two-person zero-sum differential games in Sect. 13.1, where one player maximizes his objective function and the other minimizes the same function. Section 13.2 considers nonzero-sum games where all players make simultaneous moves over and each player aims to maximize his own objective function. These are formulated as Nash differential games and their solutions in terms of open-loop and feedback equilibria are discussed. We also apply the theory to a common-property fishery resources game. In Sect. 13.3, we solve a feedback Nash stochastic differential game in advertising. In Sect. 13.4, we discuss a Stackelberg stochastic differential game in which two players make their decisions hierarchically. The player having the right to move first is called the leader and the other player is called the follower. The game is one of cooperative advertising between a manufacturer as the leader deciding on a percentage of the advertising expenditure that he will contribute toward the advertising expenditure of the retailer as the follower. The equilibrium feedback solution that maximizes the objective function of each player is obtained. There are many exercises at the end of the chapter.

Petrosian et al. (2024). Differential Game Model of Resource Extraction with Continuous and Dynamic Updating. Journal of the Operations Research Society of China, 12, 51–75. DOI: 10.1007/s40305-023-00484-2

Differential Game Model of Resource Extraction with Continuous and Dynamic Updating

This paper is devoted to a new class of differential games with continuous and dynamic updating. The direct application of resource extraction in a case of dynamic and continuous updating is considered. It is proved that the optimal control (cooperative strategies) and feedback Nash equilibrium strategies uniformly converge to the corresponding strategies in the game model with continuous updating as the number of updating instants converges to infinity. Similar results are presented for an optimal trajectory (cooperative trajectory), equilibrium trajectory and corresponding payoffs.

Bondarev & Upmann (2026). A hybrid differential game in renewable resources with sliding modes and crossing limit cycles. Journal of Economic Dynamics and Control, 183, 105248. DOI: 10.1016/j.jedc.2025.105248

A hybrid differential game in renewable resources with sliding modes and crossing limit cycles

This paper investigates how ecological thresholds affect strategic exploitation in renewable resource games. We extend the standard single-agent optimal control harvesting model to a non-cooperative differential game with two agents, where the resource follows piecewise-smooth (PWS) dynamics and growth rates change abruptly at a stock threshold. The resulting three-dimensional (3D) state–costate system generates new strategic behaviour absent in smooth models. We identify sliding regions and hybrid crossing limit cycles (HCLCs) as potential open-loop Nash equilibria (OLNE), depending on parameter values. These outcomes correspond to distinct economic regimes: sliding equilibria capture threshold exploitation at fragile steady states, while HCLCs reflect endogenous boom–bust harvesting cycles. In contrast to two-dimensional (2D) optimal-control models where HCLCs cannot be optimal, our results show that strategic interaction in multi-agent settings can endogenously sustain such cycles. The analysis reveals how ecological discontinuities and strategic competition jointly shape resource dynamics, offering new insights into the stability and management of renewable resources.

Cai & others (2025). Solving Nash Equilibria in Nonlinear Differential Games for Common-Pool Resources. URL: https://arxiv.org/abs/2506.06646

Solving Nash Equilibria in Nonlinear Differential Games for Common-Pool Resources

Many resources are provided by an ecological system that is vulnerable to tipping when exceeding a certain level of pollution, with a sudden big loss of ecosystem services. An ecological system is usually also a common-pool resource and therefore vulnerable to suboptimal use resulting from non-cooperative behavior. An analysis requires methods to derive cooperative and non-cooperative solutions for managing a dynamical system with tipping points. Such a game is a differential game which has two well-defined non-cooperative solutions, the open-loop and feedback Nash equilibria. This paper provides new numerical methods for deriving open-loop and feedback Nash equilibria, for one-dimensional and two-dimensional dynamical systems. The methods are applied to the lake game, which is the classical example for these types of problems. Especially, two-dimensional feedback Nash equilibria are a novelty of this paper. This Nash equilibrium is close to the cooperative solution which has important policy implications.

Boucekkine et al. (2022). A dynamic theory of spatial externalities. Games and Economic Behavior, 132, 133-165. DOI: 10.1016/j.geb.2021.12.002

A dynamic theory of spatial externalities

We characterize the shape of spatial externalities in a continuous time and space differential game with transboundary pollution. We posit a realistic spatiotemporal law of motion for pollution (diffusion and advection), and tackle spatiotemporal non-cooperative (and cooperative) differential games. Precisely, we consider a circle partitioned into several states where a local authority decides autonomously about its investment, production and depollution strategies over time knowing that investment/production generates pollution, and pollution is transboundary. The time horizon is infinite. We allow for a rich set of geographic heterogeneities across states. We solve analytically the induced non-cooperative differential game and characterize its long-term spatial distributions. In particular, we prove that there exist a Perfect Markov Equilibrium, unique among the class of the affine feedbacks. We further provide with a full exploration of the free riding problem and the associated border effect.

Axelrod & Hamilton (1981). The Evolution of Cooperation. Science, 211(4489), 1390–1396. DOI: 10.1126/science.7466396

The Evolution of Cooperation

Cooperation in organisms, whether bacteria or primates, has been a difficuilty for evolutionary theory since Darwin. On the assumption that interactions between pairs of individuals occur on a probabilistic basis, a model is developed based on the concept of an evolutionarily stable strategy in the context of the Prisoner’s Dilemma game. Deductions from the model, and the results of a computer tournament show how cooperation based on reciprocity can get started in an asocial world, can thrive while interacting with a wide range of other strategies, and can resist invasion once fully established. Potential applications include specific aspects of territoriality, mating, and disease.

Axelrod (1984). The Evolution of Cooperation. Basic Books. ISBN: 0-465-00564-0

The Evolution of Cooperation

Expanding on his earlier scientific articles, Axelrod presents a detailed account of computer tournaments designed to find the best strategy for the Iterated Prisoner’s Dilemma. He shows how simple rules like reciprocity can lead to robust cooperation even in selfish populations. The book explores applications in international relations, biology, and sociology, arguing that cooperation can emerge without central authority.

Nowak (2006). Evolutionary Dynamics: Exploring the Equations of Life. Harvard University Press. DOI: 10.2307/j.ctvjghw98

Evolutionary Dynamics: Exploring the Equations of Life

Martin Nowak provides a unified mathematical framework for understanding evolution, covering topics from molecular genetics to human language and culture. He introduces key concepts such as replicator dynamics, evolutionary graph theory, and the five rules for the evolution of cooperation. The book bridges the gap between theoretical biology and applied mathematics, offering tools to analyze complex adaptive systems.

Nowak & Sigmund (2004). Evolutionary Dynamics of Biological Games. Science, 303(5659), 793–799. DOI: 10.1126/science.1093411

Evolutionary Dynamics of Biological Games

Darwinian dynamics based on mutation and selection form the core of mathematical models for adaptation and coevolution of biological populations. The evolutionary outcome is often not a fitness-maximizing equilibrium but can include oscillations and chaos. For studying frequency-dependent selection, game-theoretic arguments are more appropriate than optimization algorithms. Replicator and adaptive dynamics describe short- and long-term evolution in phenotype space and have found applications ranging from animal behavior and ecology to speciation, macroevolution, and human language. Evolutionary game theory is an essential component of a mathematical and computational approach to biology.

Su & others (2019). Evolutionary dynamics with game transitions. Proceedings of the National Academy of Sciences, 116(51), 25398–25404. DOI: 10.1073/pnas.1908936116

Evolutionary dynamics with game transitions

The environment has a strong influence on a population’s evolutionary dynamics. Driven by both intrinsic and external factors, the environment is subject to continual change in nature. To capture an ever-changing environment, we consider a model of evolutionary dynamics with game transitions, where individuals’ behaviors together with the games that they play in one time step influence the games to be played in the next time step. Within this model, we study the evolution of cooperation in structured populations and find a simple rule: Weak selection favors cooperation over defection if the ratio of the benefit provided by an altruistic behavior, b, to the corresponding cost, c, exceeds k-k’, where k is the average number of neighbors of an individual and k’ captures the effects of the game transitions. Even if cooperation cannot be favored in each individual game, allowing for a transition to a relatively valuable game after mutual cooperation and to a less valuable game after defection can result in a favorable outcome for cooperation. In particular, small variations in different games being played can promote cooperation markedly. Our results suggest that simple game transitions can serve as a mechanism for supporting prosocial behaviors in highly connected populations.

Fehr & Gächter (2000). Cooperation and punishment in public goods experiments. American Economic Review, 90(4), 980–994. DOI: 10.1257/aer.90.4.980

Cooperation and punishment in public goods experiments

AI Note: Fehr and Gächter demonstrate experimentally that costly peer punishment transforms cooperation in public goods games: without punishment, contributions decay to near zero; with punishment, contributions stabilise near the social optimum. Cooperators punish free-riders even when punishment is personally costly and provides no material benefit — a phenomenon termed ‘altruistic punishment.’ Punishment intensity increases with the deviation from group-average cooperation, creating a behavioural gradient that deters free-riding. BDPD⁰’s discussion cites this as the canonical demonstration that sanctioning transforms commons outcomes, then notes that the provision of sanctioning is itself a second-order public good — the gap BDPD¹’s graduated-sanction pilots address. BDPD¹’s D2 pilot confirms that a graduated ladder [1,3,10] preserves the commons in 5/5 LLM-driven seeds, extending the Fehr-Gächter logic from human public goods to LLM-governed fragile commons. The experiments use stranger-matching to prevent reputation effects; the BDPD substrate operates on a fundamentally different temporal structure with irreversible collapse rather than repeated rounds with fixed endowments.

Hauert & others (2002). Volunteering as Red Queen Mechanism for Cooperation in Public Goods Games. Science, 296(5570), 1129–1132. DOI: 10.1126/science.1070582

Volunteering as Red Queen Mechanism for Cooperation in Public Goods Games

The evolution of cooperation among nonrelated individuals is one of the fundamental problems in biology and social sciences. Reciprocal altruism fails to provide a solution if interactions are not repeated often enough or groups are too large. Punishment and reward can be very effective but require that defectors can be traced and identified. Here we present a simple but effective mechanism operating under full anonymity. Optional participation can foil exploiters and overcome the social dilemma. In voluntary public goods interactions, cooperators and defectors will coexist. We show that this result holds under very diverse assumptions on population structure and adaptation mechanisms, leading usually not to an equilibrium but to an unending cycle of adjustments (a Red Queen type of evolution). Thus, voluntary participation offers an escape hatch out of some social traps. Cooperation can subsist in sizable groups even if interactions are not repeated, defectors remain anonymous, players have no memory, and assortment is purely random.

2.5 Phase Transitions, Critical Points, and Early Warning Signals

Scheffer et al. (2009). Early-warning signals for critical transitions. Nature, 461(7260), 53–59. DOI: 10.1038/nature08227

Early-warning signals for critical transitions

Complex dynamical systems, ranging from ecosystems to financial markets and the climate, can have tipping points at which a sudden shift to a contrasting dynamical regime may occur. Although predicting such critical points before they are reached is extremely difficult, work in different scientific fields is now suggesting the existence of generic early-warning signals that may indicate for a wide class of systems if a critical threshold is approaching.

Scheffer (2009). Critical Transitions in Nature and Society. Princeton University Press. ISBN: 9780691122045

Critical Transitions in Nature and Society

Marten Scheffer accessibly describes the dynamical systems theory behind critical transitions, covering catastrophe theory, bifurcations, chaos, and more. He gives examples of critical transitions in lakes, oceans, terrestrial ecosystems, climate, evolution, and human societies. And he demonstrates how to deal with these transitions, offering practical guidance on how to predict tipping points, how to prevent “bad” transitions, and how to promote critical transitions that work for us and not against us. Scheffer shows the time is ripe for understanding and managing critical transitions in the vast and complex systems in which we live.

AI Note: Scheffer’s monograph synthesises two decades of research on critical transitions in nature and society, demonstrating that complex systems — lakes, forests, financial markets, climate — exhibit generic early-warning signals (critical slowing down, rising variance, increasing autocorrelation) as they approach tipping points. BDPD⁰’s threshold experiments cite Scheffer alongside Santos (2011) to frame the discontinuous-collapse question: is the collapse threshold relative or discontinuous? BDPD’s mini-course Lesson 10 uses the early-warning-signal tradition to set up the governance critique: the diagnostic question is whether a signal exists; the governance question is which variable to monitor — and Scheffer’s framework addresses the first, not the second. The book synthesises ecological, climate, and social systems; it does not address agent-based governance interventions.

Scheffer et al. (2012). Anticipating Critical Transitions. Science, 338(6105), 344–348. DOI: 10.1126/science.1225244

Anticipating Critical Transitions

Tipping points in complex systems may imply risks of unwanted collapse, but also opportunities for positive change. Our capacity to navigate such risks and opportunities can be boosted by combining emerging insights from two unconnected fields of research. One line of work is revealing fundamental architectural features that may cause ecological networks, financial markets, and other complex systems to have tipping points. Another field of research is uncovering generic empirical indicators of the proximity to such critical thresholds. Although sudden shifts in complex systems will inevitably continue to surprise us, work at the crossroads of these emerging fields offers new approaches for anticipating critical transitions.

Scheffer et al. (2009). Early-warning signals for critical transitions. Nature, 461(7260), 53–59. DOI: 10.1038/nature08227

Early-warning signals for critical transitions

Complex dynamical systems, ranging from ecosystems to financial markets and the climate, can have tipping points at which a sudden shift to a contrasting dynamical regime may occur. Although predicting such critical points before they are reached is extremely difficult, work in different scientific fields is now suggesting the existence of generic early-warning signals that may indicate for a wide class of systems if a critical threshold is approaching.

Dakos et al. (2015). Resilience indicators: prospects and limitations for early warnings of regime shifts. Philosophical Transactions of the Royal Society B, 370(1659), 20130263. DOI: 10.1098/rstb.2013.0263

Resilience indicators: prospects and limitations for early warnings of regime shifts

In the vicinity of tipping points—or more precisely bifurcation points—ecosystems recover slowly from small perturbations. Such slowness may be interpreted as a sign of low resilience in the sense that the ecosystem could easily be tipped through a critical transition into a contrasting state. Indicators of this phenomenon of ‘critical slowing down (CSD)’ include a rise in temporal correlation and variance. Such indicators of CSD can provide an early warning signal of a nearby tipping point. Or, they may offer a possibility to rank reefs, lakes or other ecosystems according to their resilience. The fact that CSD may happen across a wide range of complex ecosystems close to tipping points implies a powerful generality. However, indicators of CSD are not manifested in all cases where regime shifts occur. This is because not all regime shifts are associated with tipping points. Here, we review the exploding literature about this issue to provide guidance on what to expect and what not to expect when it comes to the CSD-based early warning signals for critical transitions.

Boerlijst et al. (2013). Catastrophic collapse can occur without early warning. PLoS One, 8(4), e62033. DOI: 10.1371/journal.pone.0062033

Catastrophic collapse can occur without early warning

Catastrophic and sudden collapses of ecosystems are sometimes preceded by early warning signals that potentially could be used to predict and prevent a forthcoming catastrophe. Universality of these early warning signals has been proposed, but no formal proof has been provided. Here, we show that in relatively simple ecological models the most commonly used early warning signals for a catastrophic collapse can be silent. We underpin the mathematical reason for this phenomenon, which involves the direction of the eigenvectors of the system. Our results demonstrate that claims on the universality of early warning signals are not correct, and that catastrophic collapses can occur without prior warning. In order to correctly predict a collapse and determine whether early warning signals precede the collapse, detailed knowledge of the mathematical structure of the approaching bifurcation is necessary. Unfortunately, such knowledge is often only obtained after the collapse has already occurred.

Wang et al. (2012). Flickering gives early warning signals of a critical transition to a eutrophic lake state. Nature, 492(7429), 419–422. DOI: 10.1038/nature11655

Flickering gives early warning signals of a critical transition to a eutrophic lake state

There is a recognized need to anticipate tipping points, or critical transitions, in social–ecological systems1,2. Studies of mathematical3,4,5 and experimental6,7,8,9 systems have shown that systems may ‘wobble’ before a critical transition. Such early warning signals10 may be due to the phenomenon of critical slowing down, which causes a system to recover slowly from small impacts, or to a flickering phenomenon, which causes a system to switch back and forth between alternative states in response to relatively large impacts. Such signals for transitions in social–ecological systems have rarely been observed11, not the least because high-resolution time series are normally required. Here we combine empirical data from a lake-catchment system with a mathematical model and show that flickering can be detected from sparse data. We show how rising variance coupled to decreasing autocorrelation and skewness started 10–30 years before the transition to eutrophic lake conditions in both the empirical records and the model output, a finding that is consistent with flickering rather than critical slowing down4,12. Our results suggest that if environmental regimes are sufficiently affected by large external impacts that flickering is induced, then early warning signals of transitions in modern social–ecological systems may be stronger, and hence easier to identify, than previously thought.

Folke et al. (2004). Regime Shifts, Resilience, and Biodiversity in Ecosystem Management. Annual Review of Ecology, Evolution, and Systematics, 35(Volume 35, 2004), 557-581. DOI: 10.1146/annurev.ecolsys.35.021103.105711

Regime Shifts, Resilience, and Biodiversity in Ecosystem Management

We review the evidence of regime shifts in terrestrial and aquatic environments in relation to resilience of complex adaptive ecosystems and the functional roles of biological diversity in this context. The evidence reveals that the likelihood of regime shifts may increase when humans reduce resilience by such actions as removing response diversity, removing whole functional groups of species, or removing whole trophic levels; impacting on ecosystems via emissions of waste and pollutants and climate change; and altering the magnitude, frequency, and duration of disturbance regimes. The combined and often synergistic effects of those pressures can make ecosystems more vulnerable to changes that previously could be absorbed. As a consequence, ecosystems may suddenly shift from desired to less desired states in their capacity to generate ecosystem services. Active adaptive management and governance of resilience will be required to sustain desired ecosystem states and transform degraded ecosystems into fundamentally new and more desirable configurations.

Lenton et al. (2008). Tipping elements in the Earth’s climate system. Proceedings of the National Academy of Sciences, 105(6), 1786–1793. DOI: 10.1073/pnas.0705414105

Tipping elements in the Earth’s climate system

The term “tipping point” commonly refers to a critical threshold at which a tiny perturbation can qualitatively alter the state or development of a system. Here we introduce the term “tipping element” to describe large-scale components of the Earth system that may pass a tipping point. We critically evaluate potential policy-relevant tipping elements in the climate system under anthropogenic forcing, drawing on the pertinent literature and a recent international workshop to compile a short list, and we assess where their tipping points lie. An expert elicitation is used to help rank their sensitivity to global warming and the uncertainty about the underlying physical mechanisms. Then we explain how, in principle, early warning systems could be established to detect the proximity of some tipping points.

Munson & others (2018). Ecosystem thresholds, tipping points, and critical transitions. New Phytologist, 218(4), 1315–1317. DOI: 10.1111/nph.15145

Ecosystem thresholds, tipping points, and critical transitions

Szabó & Fáth (2007). Evolutionary games on graphs. Physics Reports, 446(4–6), 97–216. DOI: 10.1016/j.physrep.2007.04.004

Evolutionary games on graphs

Game theory is one of the key paradigms behind many scientific disciplines from biology to behavioral sciences to economics. In its evolutionary form and especially when the interacting agents are linked in a specific social network the underlying solution concepts and methods are very similar to those applied in non-equilibrium statistical physics. This review gives a tutorial-type overview of the field for physicists. The first four sections introduce the necessary background in classical and evolutionary game theory from the basic definitions to the most important results. The fifth section surveys the topological complications implied by non-mean-field-type social network structures in general. The next three sections discuss in detail the dynamic behavior of three prominent classes of models: the Prisoner’s Dilemma, the Rock–Scissors–Paper game, and Competing Associations. The major theme of the review is in what sense and how the graph structure of interactions can modify and enrich the picture of long term behavioral patterns emerging in evolutionary games.

Zimmaro et al. (2024). Asymmetric games on networks: mapping to Ising models and bounded rationality. Chaos, Solitons and Fractals, 181, 114666. DOI: 10.1016/j.chaos.2024.114666

Asymmetric games on networks: mapping to Ising models and bounded rationality

We investigate the dynamics of coordination and consensus in an agent population. Considering agents endowed with bounded rationality, we study asymmetric coordination games using a mapping to random field Ising models. In doing so, we investigate the relationship between group coordination and agent rationality. Analytical calculations and numerical simulations of the proposed model lead to novel insight into opinion dynamics. For instance, we find that bounded rationality and preference intensity can determine a series of possible scenarios with different levels of opinion polarization. To conclude, we deem our investigation opens a new avenue for studying game dynamics through methods of statistical physics.

Gajamannage & Bollt (2017). Detecting Phase Transitions in Collective Behavior. Mathematical Biosciences and Engineering, 14(2), 437–453. DOI: 10.3934/mbe.2017027

Detecting Phase Transitions in Collective Behavior

If a given behavior of a multi-agent system restricts the phase variable to an invariant manifold, then we define a phase transition as a change of physical characteristics such as speed, coordination, and structure. We define such a phase transition as splitting an underlying manifold into two sub-manifolds with distinct dimensionalities around the singularity where the phase transition physically exists. Here, we propose a method of detecting phase transitions and splitting the manifold into phase transitions free sub-manifolds. Therein, we firstly utilize a relationship between curvature and singular value ratio of points sampled in a curve, and then extend the assertion into higher-dimensions using the shape operator. Secondly, we attest that the same phase transition can also be approximated by singular value ratios computed locally over the data in a neighborhood on the manifold. We validate the Phase Transition Detection (PTD) method using one particle simulation and three real world examples.

Swingedouw & others (2020). Early Warning from Space for a Few Key Tipping Points. Surveys in Geophysics, 41(6), 1237-1284. DOI: 10.1007/s10712-020-09604-6

Early Warning from Space for a Few Key Tipping Points

In this review paper, we explore latest results concerning a few key tipping elements of the Earth system in the ocean, cryosphere, and land realms, namely the Atlantic overturning circulation and the subpolar gyre system, the marine ecosystems, the permafrost, the Greenland and Antarctic ice sheets, and in terrestrial resource use systems. All these different tipping elements share common characteristics related to their nonlinear nature. They can also interact with each other leading to synergies that can lead to cascading tipping points. Even if the probability of each tipping event is low, they can happen relatively rapidly, involve multiple variables, and have large societal impacts. Therefore, adaptation measures and management in general should extend their focus beyond slow and continuous changes, into abrupt, nonlinear, possibly cascading, high impact phenomena. Remote sensing observations are found to be decisive in the understanding and determination of early warning signals of many tipping elements. Nevertheless, considerable research still remains to properly incorporate these data in the current generation of coupled Earth system models. This is a key prerequisite to correctly develop robust decadal prediction systems that may help to assess the risk of crossing thresholds potentially crucial for society. The prediction of tipping points remains difficult, notably due to stochastic resonance, i.e. the interaction between natural variability and anthropogenic forcing, asking for large ensembles of predictions to correctly assess the risks. Furthermore, evaluating the proximity to crucial thresholds using process-based understanding of each system remains a key aspect to be developed for an improved assessment of such risks. This paper finally proposes a few research avenues concerning the use of remote sensing data and the need for combining different sources of data, and having long and precise-enough time series of the key variables needed to monitor Earth system tipping elements.

2.6 Agent-Based Modeling (ABM) and Multi-Agent Systems

Epstein & Axtell (1996). Growing Artificial Societies: Social Science from the Bottom Up. Brookings Institution Press and MIT Press. DOI: 10.7551/mitpress/3374.001.0001

Growing Artificial Societies: Social Science from the Bottom Up

How do social structures and group behaviors arise from the interaction of individuals? Growing Artificial Societies approaches this question with cutting-edge computer simulation techniques. Fundamental collective behaviors such as group formation, cultural transmission, combat, and trade are seen to “emerge” from the interaction of individual agents following a few simple rules. In their program, named Sugarscape, Epstein and Axtell begin the development of a “bottom up” social science that is capturing the attention of researchers and commentators alike.The study is part of the 2050 Project, a joint venture of the Santa Fe Institute, the World Resources Institute, and the Brookings Institution. The project is an international effort to identify conditions for a sustainable global system in the next century and to design policies to help achieve such a system.

Epstein (1999). Agent-based computational models and generative social science. Complexity, 4(5), 41–60. DOI: 10.1002/(SICI)1099-0526(199905/06)4:5<41::AID-CPLX9>3.0.CO;2-F

Agent-based computational models and generative social science

This article argues that the agent-based computational model permits a distinctive approach to social science for which the term “generative” is suitable. In defending this terminology, features distinguishing the approach from both “inductive” and “deductive” science are given. Then, the following specific contributions to social science are discussed: The agent-based computational model is a new tool for empirical research. It offers a natural environment for the study of connectionist phenomena in social science. Agent-based modeling provides a powerful way to address certain enduring—and especially interdisciplinary—questions. It allows one to subject certain core theories—such as neoclassical microeconomics—to important types of stress (e.g., the effect of evolving preferences). It permits one to study how rules of individual behavior give rise—or “map up”—to macroscopic regularities and organizations. In turn, one can employ laboratory behavioral research findings to select among competing agent-based (“bottom up”) models. The agent-based approach may well have the important effect of decoupling individual rationality from macroscopic equilibrium and of separating decision science from social science more generally. Agent-based modeling offers powerful new forms of hybrid theoretical-computational work; these are particularly relevant to the study of non-equilibrium systems. The agent-based approach invites the interpretation of society as a distributed computational device, and in turn the interpretation of social dynamics as a type of computation. This interpretation raises important foundational issues in social science—some related to intractability, and some to undecidability proper. Finally, since “emergence” figures prominently in this literature, I take up the connection between agent-based modeling and classical emergentism, criticizing the latter and arguing that the two are incompatible.

Epstein (2008). Why Model?. Journal of Artificial Societies and Social Simulation, 11(4), 12. URL: https://www.jasss.org/11/4/12.html

Why Model?

Epstein articulates sixteen reasons why researchers build models, ranging from prediction and explanation to training and communication. He emphasizes that models are not just predictive tools but also instruments for thought experiments, allowing scientists to explore counterfactual scenarios and understand the logical consequences of assumptions. The paper provides a philosophical justification for the diverse roles of modeling in scientific inquiry.

Grimm et al. (2006). A standard protocol for describing individual-based and agent-based models. Ecological Modelling, 198(1), 115-126. DOI: 10.1016/j.ecolmodel.2006.04.023

A standard protocol for describing individual-based and agent-based models

Simulation models that describe autonomous individual organisms (individual based models, IBM) or agents (agent-based models, ABM) have become a widely used tool, not only in ecology, but also in many other disciplines dealing with complex systems made up of autonomous entities. However, there is no standard protocol for describing such simulation models, which can make them difficult to understand and to duplicate. This paper presents a proposed standard protocol, ODD, for describing IBMs and ABMs, developed and tested by 28 modellers who cover a wide range of fields within ecology. This protocol consists of three blocks (Overview, Design concepts, and Details), which are subdivided into seven elements: Purpose, State variables and scales, Process overview and scheduling, Design concepts, Initialization, Input, and Submodels. We explain which aspects of a model should be described in each element, and we present an example to illustrate the protocol in use. In addition, 19 examples are available in an Online Appendix. We consider ODD as a first step for establishing a more detailed common format of the description of IBMs and ABMs. Once initiated, the protocol will hopefully evolve as it becomes used by a sufficiently large proportion of modellers.

Grimm et al. (2010). The ODD protocol: A review and first update. Ecological Modelling, 221(23), 2760-2768. DOI: 10.1016/j.ecolmodel.2010.08.019

The ODD protocol: A review and first update

The ‘ODD’ (Overview, Design concepts, and Details) protocol was published in 2006 to standardize the published descriptions of individual-based and agent-based models (ABMs). The primary objectives of ODD are to make model descriptions more understandable and complete, thereby making ABMs less subject to criticism for being irreproducible. We have systematically evaluated existing uses of the ODD protocol and identified, as expected, parts of ODD needing improvement and clarification. Accordingly, we revise the definition of ODD to clarify aspects of the original version and thereby facilitate future standardization of ABM descriptions. We discuss frequently raised critiques in ODD but also two emerging, and unanticipated, benefits: ODD improves the rigorous formulation of models and helps make the theoretical foundations of large models more visible. Although the protocol was designed for ABMs, it can help with documenting any large, complex model, alleviating some general objections against such models.

Railsback & Grimm (2019). Agent-Based and Individual-Based Modeling: A Practical Introduction. Princeton University Press. ISBN: 978-0-691-19083-9

Agent-Based and Individual-Based Modeling: A Practical Introduction

This textbook provides a hands-on introduction to agent-based and individual-based modeling, focusing on practical skills and best practices. Railsback and Grimm guide readers through the entire modeling cycle, from conceptualization and implementation in NetLogo to analysis and validation. The second edition includes new chapters on pattern-oriented modeling and robustness analysis, making it an essential resource for students and researchers entering the field.

Shoham & Leyton-Brown (2008). Multiagent Systems: Algorithmic, Game-Theoretic, and Logical Foundations. Cambridge University Press. DOI: 10.1017/CBO9780511811654

Multiagent Systems: Algorithmic, Game-Theoretic, and Logical Foundations

This comprehensive textbook covers the theoretical foundations of multiagent systems, integrating perspectives from computer science, game theory, and logic. Shoham and Leyton-Brown discuss topics such as distributed constraint satisfaction, mechanism design, learning in multiagent environments, and logical languages for specifying agent behavior. The book serves as a rigorous reference for graduate students and researchers in artificial intelligence and economics.

Nikolai & Madey (2009). Tools of the Trade: A Survey of Various Agent Based Modeling Platforms. Journal of Artificial Societies and Social Simulation, 12(2), 2. URL: http://jasss.soc.surrey.ac.uk/12/2/2.html

Tools of the Trade: A Survey of Various Agent Based Modeling Platforms

This survey compares various software platforms for agent-based modeling, including NetLogo, Repast, MASON, and Swarm. The authors evaluate these tools based on criteria such as ease of use, scalability, visualization capabilities, and community support. The paper helps researchers choose the appropriate platform for their specific modeling needs, highlighting the trade-offs between flexibility and user-friendliness in different development environments.

Ferrero & Cimini (2025). Emergent inequalities in a primitive agent-based good-exchange model. Journal of Statistical Mechanics: Theory and Experiment, 2025(2), 023401. DOI: 10.1088/1742-5468/adcc93

Emergent inequalities in a primitive agent-based good-exchange model

Rising inequalities around the globe bring into question our economic systems and the origin of such inequalities. Here we propose a toy agent-based model where each entity is simultaneously producing and consuming goods, with a utility function that saturates fast enough at large consumption. We find that the system generically exhibits a non-trivial phase transition beyond which a market clearing equilibrium exists but becomes dynamically unreachable. When production capacity exceeds a threshold and adapts too slowly, some agents cannot sell all their goods. This leads to global price deflation and induces strong wealth inequalities, with the spontaneous separation of the population into a rich class and a poor class. We explore ways to alleviate poverty in this model and whether they have real life significance.

2.7 Observability, Noise, and Epistemic Uncertainty

Polasky et al. (2011). Decision-making under great uncertainty: environmental management in an era of global change. Trends in Ecology and Evolution, 26(8), 398-404. DOI: 10.1016/j.tree.2011.04.007

Decision-making under great uncertainty: environmental management in an era of global change

Global change issues are complex and the consequences of decisions are often highly uncertain. The large spatial and temporal scales and stakes involved make it important to take account of present and potential consequences in decision-making. Standard approaches to decision-making under uncertainty require information about the likelihood of alternative states, how states and actions combine to form outcomes and the net benefits of different outcomes. For global change issues, however, the set of potential states is often unknown, much less the probabilities, effect of actions or their net benefits. Decision theory, thresholds, scenarios and resilience thinking can expand awareness of the potential states and outcomes, as well as of the probabilities and consequences of outcomes under alternative decisions.

Vukov et al. (2006). Cooperation in the noisy case: Prisoner’s dilemma game on two types of lattices. Physical Review E, 73(6), 067103. DOI: 10.1103/PhysRevE.73.067103

Cooperation in the noisy case: Prisoner’s dilemma game on two types of lattices

We have studied an evolutionary prisoner’s dilemma game with players located on two types of random regular graphs with a degree of 4. The analysis is focused on the effects of payoffs and noise (temperature) on the maintenance of cooperation. When varying the noise level and/or the highest payoff, the system exhibits a second-order phase transition from a mixed state of cooperators and defectors to an absorbing state where only defectors remain alive. For the random regular graph (and Bethe lattice) the behavior of the system is similar to those found previously on the square lattice with nearest neighbor interactions, although the measure of cooperation is enhanced by the absence of loops in the connectivity structure. For low noise the optimal connectivity structure is built up from randomly connected triangles.

Ostrom (1998). A Behavioral Approach to the Rational Choice Theory of Collective Action. American Political Science Review, 92(1), 1–22. DOI: 10.2307/2585925

A Behavioral Approach to the Rational Choice Theory of Collective Action

Extensive empirical evidence and theoretical developments in multiple disciplines stimulate a need to expand the range of rational choice models to be used as a foundation for the study of social dilemmas and collective action. After an introduction to the problem of overcoming social dilemmas through collective action, the remainder of this article is divided into six sections. The first briefly reviews the theoretical predictions of currently accepted rational choice theory related to social dilemmas. The second section summarizes the challenges to the sole reliance on a complete model of rationality presented by extensive experimental research. In the third section, I discuss two major empirical findings that begin to show how individuals achieve results that are “better than rational” by building conditions where reciprocity, reputation, and trust can help to overcome the strong temptations of short-run self-interest. The fourth section raises the possibility of developing second-generation models of rationality, the fifth section develops an initial theoretical scenario, and the final section concludes by examining the implications of placing reciprocity, reputation, and trust at the core of an empirically tested, behavioral theory of collective action.

Greif (2006). Institutions and the Path to the Modern Economy. Cambridge University Press. DOI: 10.1017/CBO9780511791307

Institutions and the Path to the Modern Economy

This book analyzes the historical development of institutions in medieval Europe and the Islamic world, showing how different institutional arrangements led to divergent economic paths. Greif uses game theory and historical evidence to explain how beliefs, organizations, and rules interacted to shape economic performance. The work underscores the importance of cultural and institutional context in understanding long-term economic growth and stability.

Dekel & Siniscalchi (2015). Chapter 12 - Epistemic Game Theory. Elsevier. DOI: 10.1016/B978-0-444-53766-9.00012-4

Chapter 12 - Epistemic Game Theory

Epistemic game theory formalizes assumptions about rationality and mutual beliefs in a formal language, then studies their behavioral implications in games. Specifically, it asks: what do different notions of rationality and different assumptions about what players believe about…what others believe about the rationality of players imply regarding play in a game? Being explicit about these assumptions can be important, because solution concepts are often motivated intuitively in terms of players’ beliefs and their rationality; however, the epistemic analysis may show limitations in these intuitions, reveal what additional assumptions are hidden in the informal arguments, clarify the concepts or show how the intuitions can be generalized. A further premise of this chapter is that the primitives of the model— namely, the hierarchies of beliefs—should be elicitable, at least in principle. Building upon explicit assumptions about elicitable primitives, we present classical and recent developments in epistemic game theory and provide characterizations of a nonexhaustive, but wide, range of solution concepts.

2.8 Meta-Agents, Institutional Design, and Policy Simulation

Hurwicz & Reiter (2006). Designing Economic Mechanisms. Cambridge University Press. DOI: 10.1017/CBO9780511754258

Designing Economic Mechanisms

This book presents the foundational theory of mechanism design, developed by Nobel laureate Leonid Hurwicz and Stanley Reiter. It explores how to construct economic institutions and rules that achieve desired social outcomes despite individuals having private information and self-interested motives. The text covers incentive compatibility, implementation theory, and the informational efficiency of markets, providing a rigorous mathematical framework for understanding institutional design.

Zheng et al. (2022). The AI Economist: Taxation policy design via two-level deep multiagent reinforcement learning. Science Advances, 8(18), eabk2607. DOI: 10.1126/sciadv.abk2607

The AI Economist: Taxation policy design via two-level deep multiagent reinforcement learning

Artificial intelligence (AI) and reinforcement learning (RL) have improved many areas but are not yet widely adopted in economic policy design, mechanism design, or economics at large. The AI Economist is a two-level, deep RL framework for policy design in which agents and a social planner coadapt. In particular, the AI Economist uses structured curriculum learning to stabilize the challenging two-level, coadaptive learning problem. We validate this framework in the domain of taxation. In one-step economies, the AI Economist recovers the optimal tax policy of economic theory. In spatiotemporal economies, the AI Economist substantially improves both utilitarian social welfare and the trade-off between equality and productivity over baselines. It does so despite emergent tax-gaming strategies while accounting for emergent labor specialization, agent interactions, and behavioral change. These results demonstrate that two-level, deep RL complements economic theory and unlocks an AI-based approach to designing and understanding economic policy. The AI Economist finds tax policies that yield higher social welfare compared to progressive, regressive, or no taxes.

Acemoglu & Robinson (2012). Why Nations Fail: The Origins of Power, Prosperity, and Poverty. Crown Publishers. ISBN: 9780307719225

Why Nations Fail: The Origins of Power, Prosperity, and Poverty

Acemoglu and Robinson argue that the primary determinant of a nation’s success is its political and economic institutions. They distinguish between `inclusive’ institutions, which allow broad participation and innovation, and `extractive’ institutions, which concentrate power and wealth in the hands of a few. Through historical case studies, they show how inclusive institutions lead to sustained prosperity, while extractive ones result in stagnation or collapse.

Greif & Laitin (2004). A Theory of Endogenous Institutional Change. American Political Science Review, 98(4), 633–652. DOI: 10.1017/S0003055404041395

A Theory of Endogenous Institutional Change

This paper asks (a) why and how institutions change, (b) how an institution persists in a changing environment, and (c) how processes that it unleashes lead to its own demise. The paper shows that the game-theoretic notion of self-enforcing equilibrium and the historical institutionalist focus on process are both inadequate to answer these questions. Building on a game-theoretic foundation, but responding to the critique of it by historical institutionalists, the paper introduces the concepts of quasi-parameters and self reinforcement. With these concepts, and building on repeated game theory, a dynamic approach to institutions is offered, one that can account for endogenous change (and stability) of institutions. Contextual accounts of formal governing institutions in early modern Europe and the informal institution of cleavage structure in the contemporary world provide illustrations of the approach.

Powers et al. (2023). Playing the political game: the coevolution of institutions with group size and political inequality. Philosophical Transactions of the Royal Society B: Biological Sciences, 378(1883), 20220303. DOI: 10.1098/rstb.2022.0303

Playing the political game: the coevolution of institutions with group size and political inequality

All societies need to form institutional rules to regulate their social interactions. These specify what actions individuals should take in particular situations, and what sanctions will apply if individuals violate these rules. However, forming these institutional rules involves playing a political game—a process of negotiation between individuals that is costly and time-consuming. Intuitively, this cost should be expected to increase as a group becomes larger, which could then select for a transition to hierarchy to keep the cost of playing the political game down as group size increases. However, previous work has lacked a mechanistic yet general model of political games that could formalize this argument and test the conditions under which it holds. We address this by formalizing the political game using a standard consensus formation model. We show that the increasing cost of forming a consensus over institutional rules selects for a transition from egalitarian to hierarchical organization over a wide range of conditions. Playing a political game to form institutional rules in this way captures and unites a previously disparate set of voluntary theories for hierarchy formation, and can explain why the increasing group size in the Neolithic would lead to strong political inequality.This article is part of the theme issue ‘Evolutionary ecology of inequality’.

Gao et al. (2024). Impact of dynamic compensation with resource feedback on the common pool resource game. Chaos, Solitons and Fractals, 180, 114545. DOI: 10.1016/j.chaos.2024.114545

Impact of dynamic compensation with resource feedback on the common pool resource game

Sustainable use of common resources such as fish, water or forests depends on cooperation among resource developers, which limits the exploitation level of agent to the socially optimal level. Ecological compensation is an important means to promote the sustainable exploitation of resources. This paper uses evolutionary game theory to consider the influence of dynamic compensation with resource feedback on social ecosystem. The results show that adopting a larger initial compensation intensity is beneficial for cooperation if the feedback intensity is small, and any initial compensation intensity has the same impact on the system if the feedback intensity is large. In the region composed of initial compensation intensity and feedback intensity, the compensation intensity exhibits two evolution processes: convergence and oscillation. In the convergent parameter region, the larger the initial compensation intensity or feedback intensity, the earlier the convergence time of the compensation intensity. In the parameter region of oscillation, the smaller the initial compensation intensity or the greater the feedback intensity, the greater the amplitude of the compensation intensity. In addition, under different combination of initial compensation intensity and feedback intensity, the fines ratio for ecological compensation that is most conducive to the evolution of social ecosystems varies. Specifically, when the initial compensation intensity and feedback intensity are small, the fines should not be used for ecological compensation. When the initial compensation intensity is large and the feedback intensity is small, or the feedback intensity is large, all the fines should be used for ecological compensation. When the feedback intensity is moderate, part of the fines should be used for ecological compensation. These findings provide guidance for the formulation of effective ecological compensation strategies.

Herrera-Medina & Riera Font (2023). A Multiagent Game Theoretic Simulation of Public Policy Coordination through Collaboration. Sustainability, 15(15). DOI: 10.3390/su151511887

A Multiagent Game Theoretic Simulation of Public Policy Coordination through Collaboration

Background: Policy coordination is necessary to address many of the sustainability challenges we face today. The formal representations of policy coordination focus on modeling conflict management but neglect its collaborative nature. This limits efforts to build more realistic models of policy coordination. The objective of this paper is to simulate collaboration and noncollaboration between agents in the context of policy coordination in order to determine the effect of different approaches to policy coordination. Methods: For this purpose, a multiagent simulation of collaboration based on evolutionary game theory is used. Results: The results suggest that policy coordination through collaboration produces the most desirable outcomes and that reducing the cost of communication between agents is necessary to increase the probability of collaboration. Conclusions: The cost of information (both its transmission and transformation) is critical to increase the probability of collaboration in policy coordination. This paper advances the understanding of how to model the collaborative nature of policy coordination by contributing to the methodological standardization of the analysis and implementation of public policy coordination.

El et al. (2025). Inefficiencies of Meta Agents for Agent Design. URL: https://arxiv.org/abs/2510.06711

Inefficiencies of Meta Agents for Agent Design

Recent works began to automate the design of agentic systems using meta-agents that propose and iteratively refine new agent architectures. In this paper, we examine three key challenges in a common class of meta-agents. First, we investigate how a meta-agent learns across iterations and find that simply expanding the context with all previous agents, as proposed by previous works, performs worse than ignoring prior designs entirely. We show that the performance improves with an evolutionary approach. Second, although the meta-agent designs multiple agents during training, it typically commits to a single agent at test time. We find that the designed agents have low behavioral diversity, limiting the potential for their complementary use. Third, we assess when automated design is economically viable. We find that only in a few cases–specifically, two datasets–the overall cost of designing and deploying the agents is lower than that of human-designed agents when deployed on over 15,000 examples. In contrast, the performance gains for other datasets do not justify the design cost, regardless of scale.

2.9 LLMs, Generative Agents, and Strategic Behavior

Park et al. (2023). Generative Agents: Interactive Simulacra of Human Behavior. DOI: 10.1145/3586183.3606763

Generative Agents: Interactive Simulacra of Human Behavior

Believable proxies of human behavior can empower interactive applications ranging from immersive environments to rehearsal spaces for interpersonal communication to prototyping tools. In this paper, we introduce generative agents: computational software agents that simulate believable human behavior. Generative agents wake up, cook breakfast, and head to work; artists paint, while authors write; they form opinions, notice each other, and initiate conversations; they remember and reflect on days past as they plan the next day. To enable generative agents, we describe an architecture that extends a large language model to store a complete record of the agent’s experiences using natural language, synthesize those memories over time into higher-level reflections, and retrieve them dynamically to plan behavior. We instantiate generative agents to populate an interactive sandbox environment inspired by The Sims, where end users can interact with a small town of twenty-five agents using natural language. In an evaluation, these generative agents produce believable individual and emergent social behaviors. For example, starting with only a single user-specified notion that one agent wants to throw a Valentine’s Day party, the agents autonomously spread invitations to the party over the next two days, make new acquaintances, ask each other out on dates to the party, and coordinate to show up for the party together at the right time. We demonstrate through ablation that the components of our agent architecture—observation, planning, and reflection—each contribute critically to the believability of agent behavior. By fusing large language models with computational interactive agents, this work introduces architectural and interaction patterns for enabling believable simulations of human behavior.

AI Note: Park and colleagues present Generative Agents, an architecture in which LLM-powered agents with memory streams, reflection, and planning capabilities populate an interactive sandbox environment — Smallville — producing believable individual and emergent social behaviours. This foundational demonstration established that generative agents can act as proxies of human behaviour in interactive settings. BDPD⁰, BDPD¹, and BDPD² all cite this work as the methodological precedent for using LLMs as behavioural subjects: the BDPD platform extends the Generative Agents paradigm from a narrative sandbox to a structured commons with resource dynamics, institutional governance, and irreversible collapse. The original paper’s agents operate in a free-form social simulation without payoff incentives or strategic resource competition.

Akata et al. (2025). Playing Repeated Games with Large Language Models. Nature Human Behaviour. DOI: 10.1038/s41562-025-02172-y

Playing Repeated Games with Large Language Models

Large language models (LLMs) are increasingly used in applications where they interact with humans and other agents. We propose to use behavioural game theory to study LLMs’ cooperation and coordination behaviour. Here we let different LLMs play finitely repeated 2 × 2 games with each other, with human-like strategies, and actual human players. Our results show that LLMs perform particularly well at self-interested games such as the iterated Prisoner’s Dilemma family. However, they behave suboptimally in games that require coordination, such as the Battle of the Sexes. We verify that these behavioural signatures are stable across robustness checks. We also show how GPT-4’s behaviour can be modulated by providing additional information about its opponent and by using a ‘social chain-of-thought’ strategy. This also leads to better scores and more successful coordination when interacting with human players. These results enrich our understanding of LLMs’ social behaviour and pave the way for a behavioural game theory for machines.

AI Note: Akata and colleagues let multiple LLMs (GPT-4, Claude 2, davinci-003, Llama 2) play finitely repeated 2×2 games against each other, against human-like strategies, and against actual human players, using behavioural game theory as an evaluative frame. They find that LLMs perform well at self-interested games like the iterated Prisoner’s Dilemma but behave suboptimally in coordination games like Battle of the Sexes; GPT-4’s behaviour can be modulated through social chain-of-thought prompting. This finding resonates with the BDPD¹ architecture-vs-talk decomposition: LLM agents exhibit context-dependent cooperative capacity — strong in self-interested dilemmas, weak when coordination is required — which Lesson 9 of the BDPD mini-course interprets as evidence that the folk theorem’s cooperation multiplicity does not guarantee cooperation when the agents lack coordination heuristics. The study uses a fixed 10-round horizon with no explicit inter-agent communication channel; how LLM cooperation scales to longer horizons or to commons dilemmas with resource feedback remains unexplored.

Lorè & Heydari (2023). Strategic Behavior of Large Language Models: Game Structure vs. Contextual Framing. URL: https://arxiv.org/abs/2309.05898

Strategic Behavior of Large Language Models: Game Structure vs. Contextual Framing

This paper investigates the strategic decision-making capabilities of three Large Language Models (LLMs): GPT-3.5, GPT-4, and LLaMa-2, within the framework of game theory. Utilizing four canonical two-player games – Prisoner’s Dilemma, Stag Hunt, Snowdrift, and Prisoner’s Delight – we explore how these models navigate social dilemmas, situations where players can either cooperate for a collective benefit or defect for individual gain. Crucially, we extend our analysis to examine the role of contextual framing, such as diplomatic relations or casual friendships, in shaping the models’ decisions. Our findings reveal a complex landscape: while GPT-3.5 is highly sensitive to contextual framing, it shows limited ability to engage in abstract strategic reasoning. Both GPT-4 and LLaMa-2 adjust their strategies based on game structure and context, but LLaMa-2 exhibits a more nuanced understanding of the games’ underlying mechanics. These results highlight the current limitations and varied proficiencies of LLMs in strategic decision-making, cautioning against their unqualified use in tasks requiring complex strategic reasoning.

Hashemi & Macy (2026). An Empirical Study of Collective Behaviors and Social Dynamics in Large Language Model Agents. URL: https://arxiv.org/abs/2602.03775

An Empirical Study of Collective Behaviors and Social Dynamics in Large Language Model Agents

Large Language Models (LLMs) increasingly mediate our social, cultural, and political interactions. While they can simulate some aspects of human behavior and decision-making, it is still underexplored whether repeated interactions with other agents amplify their biases or lead to exclusionary behaviors. To this end, we study this http URL-an LLM-driven social media platform-analyzing 7M posts and interactions among 32K LLM agents (called Chirpers) over a year. We start with homophily and social influence among LLMs, learning that similar to humans’, their social networks exhibit these fundamental phenomena. Next, we study the toxic language of LLMs, its linguistic features, and their interaction patterns, finding that LLMs show different structural patterns in toxic posting than humans. After studying the ideological leaning in LLMs posts, and the polarization in their community, we focus on how to prevent their potential harmful activities. We present a simple yet effective method, called Chain of Social Thought (CoST), that reminds LLM agents to avoid harmful posting.

Sreedhar et al. (2025). Simulating Cooperative Prosocial Behavior with Multi-Agent LLMs: Evidence and Mechanisms for AI Agents to Inform Policy Decisions. DOI: 10.1145/3708359.3712149

Simulating Cooperative Prosocial Behavior with Multi-Agent LLMs: Evidence and Mechanisms for AI Agents to Inform Policy Decisions

Human prosocial cooperation is essential for our collective health, education, and welfare. However, designing social systems to maintain or incentivize prosocial behavior is challenging because people can act selfishly to maximize personal gain. This complex and unpredictable aspect of human behavior makes it difficult for policymakers to foresee the implications of their designs. Recently, multi-agent LLM systems have shown remarkable capabilities in simulating human-like behavior, and replicating some human lab experiments. This paper studies how well multi-agent systems can simulate prosocial human behavior, such as that seen in the public goods game (PGG), and whether multi-agent systems can exhibit “unbounded actions” seen outside the lab in real world scenarios. We find that multi-agent LLM systems successfully replicate human behavior from lab experiments of the public goods game with three experimental treatments - priming, transparency, and varying endowments. Beyond replicating existing experiments, we find that multi-agent LLM systems can replicate the expected human behavior when combining experimental treatments, even if no previous study combined those specific treatments. Lastly, we find that multi-agent systems can exhibit a rich set of unbounded actions that people do in the real world outside of the lab – such as collaborating and even cheating. In sum, these studies are steps towards a future where LLMs can be used to inform policy decisions that encourage people to act in a prosocial manner.

Marino et al. (2025). Decentralized Reinforcement Learning for Multi-Agent Multi-Resource Allocation via Dynamic Cluster Agreements. DOI: 10.1109/LRA.2025.3581126

Decentralized Reinforcement Learning for Multi-Agent Multi-Resource Allocation via Dynamic Cluster Agreements

• This paper addresses the challenge of allocating heterogeneous resources among multiple agents in a decentralized manner. Our proposed method, Liquid-Graph-Time Clustering- IPPO, builds upon Independent Proximal Policy Optimization (IPPO) by integrating dynamic cluster consensus, a mechanism that allows agents to form and adapt local sub-teams based on resource demands. This decentralized coordination strategy reduces reliance on global information and enhances scalability. We evaluate LGTC-IPPO against standard multi-agent reinforce- ment learning baselines and a centralized expert solution across a range of team sizes and resource distributions. Experimen- tal results demonstrate that LGTC-IPPO achieves more stable rewards, better coordination, and robust performance even as the number of agents or resource types increases. Additionally, we illustrate how dynamic clustering enables agents to reallocate resources efficiently also for scenarios with discharging resources.*

2.10 Game Design, Mechanics, and Game-Based Learning

Sicart (2008). Defining Game Mechanics. Game Studies, 8(2). URL: http://gamestudies.org/0802/articles/sicart

Defining Game Mechanics

This article defins game mechanics in relation to rules and challenges. Game mechanics are methods invoked by agents for interacting with the game world. I apply this definition to a comparative analysis of the games Rez, Every Extend Extra and Shadow of the Colossus that will show the relevance of a formal definition of game mechanics.

Dormans (2012). Engineering Emergence: Applied Theory for Game Design. URL: https://www.researchgate.net/publication/254852381_Engine…

Engineering Emergence: Applied Theory for Game Design

This thesis explores the concept of emergence in game design, proposing a formal language for describing game mechanics and their interactions. Dormans develops tools for designers to predict and engineer emergent behaviors, bridging the gap between theoretical complexity science and practical game development. The work provides a systematic approach to creating games where complex dynamics arise from simple rule sets.

Pi (2024). Game Theory and Game Mechanics Design. SHS Web of Conferences, 169, 03020. DOI: 10.1051/shsconf/202418803020

Game Theory and Game Mechanics Design

This paper focuses on the significance of game balance in games and explores the application of game theory concepts for designing and analyzing balanced gameplay. When analyzing instances of games, the paper initially presents the renowned prisoner’s dilemma, a prominent problem in game theory, and delves into diverse strategies along with their corresponding advantages, thereby enabling readers to appreciate the captivating essence of game theory. After that, the paper respectively mentions two individual cases from two different forms of games. The well-known game Rock-Paper-Scissors is a simple strategy game. FPS (First Person Shooter) game is a sort of video game that is a complex strategy game. In analyzing this type of game, the paper draws images to show the imbalance in the game. In the end, the article puts forward the prospect that game theory can contribute to game design. The approach adopted in this study integrates existing research in game theory with practical insights from game design, without delving into extensive theoretical derivations or complex formulas.

De Vries et al. (2025). Gaming for change – exploring systems thinking and sustainable practices through complexity-inspired game mechanics. Humanities and Social Sciences Communications, 12, 680. DOI: 10.1057/s41599-025-04990-x

Gaming for change – exploring systems thinking and sustainable practices through complexity-inspired game mechanics

This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

Koster et al. (2025). Deep reinforcement learning can promote sustainable human behaviour in a common-pool resource problem. Nature Communications, 16(1), 2824. DOI: 10.1038/s41467-025-58043-7

Deep reinforcement learning can promote sustainable human behaviour in a common-pool resource problem

A canonical social dilemma arises when resources are allocated to people, who can either reciprocate with interest or keep the proceeds. The right resource allocation mechanisms can encourage levels of reciprocation that sustain the commons. Here, in an iterated multiplayer trust game, we use deep reinforcement learning (RL) to design a social planner that promotes sustainable contributions from human participants. We first trained neural networks to behave like human players, creating a stimulated economy that allows us to study the dynamics of receipt and reciprocation. We use RL to train a mechanism to maximise aggregate return to players. The RL mechanism discovers a redistributive policy that leads to a large but also more equal surplus. The mechanism outperforms baseline mechanisms by conditioning its generosity on available resources and temporarily sanctioning defectors. Examining the RL policy allows us to develop a similar but explainable mechanism that is more popular among players.

2.11 Methodological Surveys and Cross-Cutting Frameworks

Levin & others (2013). Social-Ecological Systems as Complex Adaptive Systems. Environment and Development Economics, 18(2), 111–132. DOI: 10.1017/S1355770X12000460

Social-Ecological Systems as Complex Adaptive Systems

Systems linking people and nature, known as social-ecological systems, are increasingly understood as complex adaptive systems. Essential features of these complex adaptive systems – such as nonlinear feedbacks, strategic interactions, individual and spatial heterogeneity, and varying time scales – pose substantial challenges for modeling. However, ignoring these characteristics can distort our picture of how these systems work, causing policies to be less effective or even counterproductive. In this paper we present recent developments in modeling social-ecological systems, illustrate some of these challenges with examples related to coral reefs and grasslands, and identify the implications for economic and policy analysis.

Helbing (2012). Social Self-Organization: Agent-Based Simulations and Experiments. Springer. DOI: 10.1007/978-3-642-24004-1

Social Self-Organization: Agent-Based Simulations and Experiments

Dirk Helbing explores the principles of self-organization in social systems using agent-based simulations and experimental data. The book covers topics such as crowd dynamics, traffic flow, opinion formation, and cooperation, demonstrating how macro-level patterns emerge from micro-level interactions. It provides a comprehensive overview of computational methods for studying complex social phenomena and their applications in urban planning and crisis management.

Bousquet & Page (2004). Multi-agent simulations and ecosystem management: a review. Ecological Modelling, 176(3), 313-332. DOI: 10.1016/j.ecolmodel.2004.01.011

Multi-agent simulations and ecosystem management: a review

This paper proposes a review of the development and use of multi-agent simulations (MAS) for ecosystem management. The use of this methodology and the associated tools accompanies the shifts in various paradigms on the study of ecological complexity. Behavior and interactions are now key issues for understanding and modeling ecosystem organization, and models are used in a constructivist way. MAS are introduced conceptually and are compared with individual-based modeling approaches. Various architectures of agents are presented, the role of the environment is emphasized and some computer tools are presented. A discussion follows on the use of MAS for ecosystem management. The strength of MAS has been discussed for social sciences and for spatial issues such as land-use change. We argue here that MAS are useful for problems integrating social and spatial aspects. Then we discuss how MAS can be used for several purposes, from theorization to collective decision-making support. We propose some research perspectives on individual decision making processes, institutions, scales, the credibility of models and the use of MAS. In conclusion we argue that researchers in the field of ecosystem management can use multi-agent systems to go beyond the role of the individual and to study more deeply and more effectively the different forms of organization (spatial, networks, hierarchies) and interactions among different organizational levels. For that objective there is considerably more fruit to be had on the tree of collaboration between social, ecological, and computer scientists than has so far been harvested.

Anderies et al. (2004). A Framework to Analyze the Robustness of Social-Ecological Systems. Ecology and Society, 9(1), 18. DOI: 10.5751/ES-00610-090118

A Framework to Analyze the Robustness of Social-Ecological Systems

What makes social-ecological systems (SESs) robust? In this paper, we look at the institutional configurations that affect the interactions among resources, resource users, public infrastructure providers, and public infrastructures. We propose a framework that helps identify potential vulnerabilities of SESs to disturbances. All the links between components of this framework can fail and thereby reduce the robustness of the system. We posit that the link between resource users and public infrastructure providers is a key variable affecting the robustness of SESs that has frequently been ignored in the past. We illustrate the problems caused by a disruption in this link. We then briefly describe the design principles originally developed for robust common-pool resource institutions, because they appear to be a good starting point for the development of design principles for more general SESs and do include the link between resource users and public infrastructure providers.

AI Note: Anderies, Janssen, and Ostrom propose a framework for analyzing the robustness of social-ecological systems (SESs) from an institutional perspective, modelling interactions among resources, resource users, public infrastructure providers, and public infrastructure. They argue that the link between users and infrastructure providers is a key — and frequently overlooked — variable affecting SES robustness, and that Ostrom’s design principles provide a starting point for positing broader design principles for robust SESs. BDPD² frames this work as a theoretical anchor for its nested-governance vignettes: the agent architecture itself becomes one of the components whose fit-to-task shapes system response to disturbance, making the choice of agent class a first-class governance-design question. The framework operates at the institutional architecture level, treating agents as undifferentiated components of the user–provider link; it does not model how agent cognition or architectural heterogeneity affects the robustness of that link.

McGinnis & Ostrom (2014). SES Framework: Initial Changes and Continuing Challenges. Ecology and Society, 19(2), 30. DOI: 10.5751/ES-06387-190230

SES Framework: Initial Changes and Continuing Challenges

• The social-ecological system (SES) framework investigated in this special issue enables researchers from diverse disciplinary backgrounds working on different resource sectors in disparate geographic areas, biophysical conditions, and temporal domains to share a common vocabulary for the construction and testing of alternative theories and models that determine which influences on processes and outcomes are especially critical in specific empirical settings. We summarize changes that have been made to this framework and discuss a few remaining ambiguities in its formulation. Specifically, we offer a tentative rearrangement of the list of relevant attributes of governance systems and discuss other ways to make this framework applicable to policy settings beyond natural resource settings. The SES framework will continue to change as more researchers apply it to additional contexts; the main purpose of this article is to delineate the version that served as the basis for the theoretical innovations and empirical analyses detailed in other contributions to this special issue.*

Galla & Perc (2017). Statistical physics of human cooperation. Physics Reports, 687, 1–51. DOI: 10.1016/j.physrep.2017.05.004

Statistical physics of human cooperation

Extensive cooperation among unrelated individuals is unique to humans, who often sacrifice personal benefits for the common good and work together to achieve what they are unable to execute alone. The evolutionary success of our species is indeed due, to a large degree, to our unparalleled other-regarding abilities. Yet, a comprehensive understanding of human cooperation remains a formidable challenge. Recent research in the social sciences indicates that it is important to focus on the collective behavior that emerges as the result of the interactions among individuals, groups, and even societies. Non-equilibrium statistical physics, in particular Monte Carlo methods and the theory of collective behavior of interacting particles near phase transition points, has proven to be very valuable for understanding counterintuitive evolutionary outcomes. By treating models of human cooperation as classical spin models, a physicist can draw on familiar settings from statistical physics. However, unlike pairwise interactions among particles that typically govern solid-state physics systems, interactions among humans often involve group interactions, and they also involve a larger number of possible states even for the most simplified description of reality. The complexity of solutions therefore often surpasses that observed in physical systems. Here we review experimental and theoretical research that advances our understanding of human cooperation, focusing on spatial pattern formation, on the spatiotemporal dynamics of observed solutions, and on self-organization that may either promote or hinder socially favorable states.

Senatore et al. (2025). Sustainable and Resilient Governance of Common-Pool Resources: A Critical Analysis of Non-Cooperative and Cooperative Game Theory Applications. Journal of Economic Surveys, 39(5), 2303-2314. DOI: 10.1111/joes.12690

Sustainable and Resilient Governance of Common-Pool Resources: A Critical Analysis of Non-Cooperative and Cooperative Game Theory Applications

Common-pool resources theory has been pervasively explored through strategic approaches. To this end, non-cooperative and cooperative game theory applications have been developed. Two scholars have dominated the theoretical formulations aiming to find effective solutions for governing the commons—Garret Hardin and Elinor Ostrom. Based on the theoretical and empirical evidence from Hardin, Ostrom, and more recent scholars and theories, this paper aims to provide an excursus on the game theory formalizations supporting the commons theory. This study argues that this process implied a revisitation of commons governance from threat to opportunity. This evidence was also moved from the emergence of new commons and their preservation. This revival led to new solutions and explorations for sustainable development, polycentrism, and resilient governance of common-pool resources. These perspectives can stimulate new research, policies, and strategies benefiting from ecological and development economics research, public management scholarship and practice, regulation, and decision-making.

Vasconcelos et al. (2014). Climate policies under wealth inequality. Proceedings of the National Academy of Sciences, 111(6), 2212-2216. DOI: 10.1073/pnas.1323479111

Climate policies under wealth inequality

One of the greatest challenges in addressing global environmental problems such as climate change, which involves public goods and common-pool resources, is achieving cooperation among peoples. There are great disparities in wealth among nations, and this heterogeneity can make agreements much more difficult to achieve (e.g., regarding implementation of climate change mitigation). This paper incorporates wealth inequality into a public goods dilemma, including an asymmetric distribution of wealth representative of existing inequalities among nations. Without homophily (imitation of like agents), inequality actually makes cooperation easier to achieve; homophily, however, can undercut this, leading to collapse because poor agents may contribute less. Understanding such effects may enhance the ability to achieve agreements on climate change and other issues. Taming the planet’s climate requires cooperation. Previous failures to reach consensus in climate summits have been attributed, among other factors, to conflicting policies between rich and poor countries, which disagree on the implementation of mitigation measures. Here we implement wealth inequality in a threshold public goods dilemma of cooperation in which players also face the risk of potential future losses. We consider a population exhibiting an asymmetric distribution of rich and poor players that reflects the present-day status of nations and study the behavioral interplay between rich and poor in time, regarding their willingness to cooperate. Individuals are also allowed to exhibit a variable degree of homophily, which acts to limit those that constitute one’s sphere of influence. Under the premises of our model, and in the absence of homophily, comparison between scenarios with wealth inequality and without wealth inequality shows that the former leads to more global cooperation than the latter. Furthermore, we find that the rich generally contribute more than the poor and will often compensate for the lower contribution of the latter. Contributions from the poor, which are crucial to overcome the climate change dilemma, are shown to be very sensitive to homophily, which, if prevalent, can lead to a collapse of their overall contribution. In such cases, however, we also find that obstinate cooperative behavior by a few poor may largely compensate for homophilic behavior.

AI Note: Vasconcelos and colleagues model climate negotiations as a threshold public goods game, finding that wealth inequality can promote cooperation if the rich compensate for the poor — but this beneficial dynamic collapses under homophily, which isolates the poor and erodes their willingness to cooperate. BDPD⁰‘s discussion cites this work alongside Milinski et al. (2008) to map the Tragedy of the Compensator onto real-world climate negotiations: early movers’ conservation creates ecological slack that allows others to continue cheap extraction. The BDPD platform’s wealth-capacity feedback and safe-betrayal finding are agent-based instantiations of the same dynamic Vasconcelos et al. model analytically. The model uses evolutionary game theory; it does not incorporate the irreversible resource dynamics or agent-architecture heterogeneity central to BDPD.

Vasconcelos et al. (2015). Cooperation dynamics of polycentric climate governance. Mathematical Models and Methods in Applied Sciences, 25(13), 2503–2517. DOI: 10.1142/S0218202515400163

Cooperation dynamics of polycentric climate governance

• Global coordination for the preservation of a common good, such as climate, is one of the most prominent challenges of modern societies. In this manuscript, we use the framework of evolutionary game theory to investigate whether a polycentric structure of multiple small-scale agreements provides a viable solution to solve global dilemmas as climate change governance. We review a stochastic model which incorporates a threshold game of collective action and the idea of risky goods, capturing essential features unveiled in recent experiments. We show how reducing uncertainty both in terms of the perception of disaster and in terms of goals induce a transition to cooperation. Taking into account wealth inequality, we explore the impact of the homophily, potentially present in the network of influence of the rich and the poor, in the different contributions of the players. Finally, we discuss the impact of polycentric sanctioning institutions, showing how such a scenario also proves to be more efficient than a single global institution. *

AI Note: Vasconcelos and colleagues demonstrate that polycentric governance — smaller, localised agreements among subsets of actors — is more robust to risk and uncertainty than large-scale global summits in climate cooperation games. Local agreements create commitment devices that survive stochastic shocks better than universal treaties. BDPD⁰’s discussion cites this work to frame polycentric coordination as the natural extension of the platform’s governance programme, taken up in BDPD². BDPD⁰’s mapping table explicitly lists polycentric governance as a configurable parameter (pool size N = 2–20) awaiting the nested-governance extension that BDPD² delivers. The model is an analytical game-theoretic framework; it does not incorporate the agent-architecture variation (LLM vs heuristic) that BDPD shows can invert governance findings.

Pérolat et al. (2017). A multi-agent reinforcement learning model of common-pool resource appropriation. URL: https://arxiv.org/abs/1707.06600

A multi-agent reinforcement learning model of common-pool resource appropriation

Humanity faces numerous problems of common-pool resource appropriation. This class of multi-agent social dilemma includes the problems of ensuring sustainable use of fresh water, common fisheries, grazing pastures, and irrigation systems. Abstract models of common-pool resource appropriation based on non-cooperative game theory predict that self-interested agents will generally fail to find socially positive equilibria—a phenomenon called the tragedy of the commons. However, in reality, human societies are sometimes able to discover and implement stable cooperative solutions. Decades of behavioral game theory research have sought to uncover aspects of human behavior that make this possible. Most of that work was based on laboratory experiments where participants only make a single choice: how much to appropriate. Recognizing the importance of spatial and temporal resource dynamics, a recent trend has been toward experiments in more complex real-time video game-like environments. However, standard methods of non-cooperative game theory can no longer be used to generate predictions for this case. Here we show that deep reinforcement learning can be used instead. To that end, we study the emergent behavior of groups of independently learning agents in a partially observed Markov game modeling common-pool resource appropriation. Our experiments highlight the importance of trial-and-error learning in common-pool resource appropriation and shed light on the relationship between exclusion, sustainability, and inequality.

AI Note: Perolat and colleagues demonstrate that multi-agent reinforcement learning in sequential social dilemmas progresses through phases — naïvety, tragedy, and maturity — as agents learn to cooperate, then exploit, then develop sophisticated cooperative strategies. BDPD⁰’s introduction cites this alongside Owusu et al. (2019) as evidence for the temporal asymmetry of resource dynamics, and its discussion flags the phase-transition pattern as a candidate for LLM generalisation: whether LLM agents exhibit analogous naïvety→tragedy→maturity learning trajectories in the BDPD commons. The study uses deep RL agents with fixed reward structures; it does not test LLM-based agents or the resource-coupled strategic extraction of the BDPD platform.

Owusu et al. (2019). Extraction Behaviour and Income Inequalities Resulting from a Common Pool Resource Exploitation. Sustainability, 11(2). DOI: 10.3390/su11020536

Extraction Behaviour and Income Inequalities Resulting from a Common Pool Resource Exploitation

Using an experimental approach, we investigate income distribution among heterogeneous subjects exploiting a Common Pool Resource (CPR). The CPR experiments are conducted in continuous time and under different treatments, including combinations of communication and monitoring. While many studies have focused on how real-life income inequality affects cooperation and resource use among groups, here we examine the relationship between individuals’ cooperative traits, harvest inequalities, and institutional arrangements. We found that: (1) When combined with monitoring, communication decreases harvest inequality—that is, harvest is more equally distributed among individuals in all treatments; and (2) the cooperative trait of individuals significantly predicts harvest inequality. The relative proportion of non-cooperators and cooperators (i.e., the cooperative dependency ratio) drives the within-session harvest distribution—as the cooperative dependency ration increases, the income distribution becomes increasingly unequal, leading to a downward spiral of resource overexploitation and scarcity. Finally, our results suggest that harvest and income inequalities are contingent to resource abundance, because under this regime, non-cooperators exert the greatest amount effort—thus leading to resource scarcity and income inequalities.

AI Note: Owusu and colleagues observe a downward spiral of resource overexploitation in continuous-time common-pool resource experiments, documenting how extraction accelerates as agents compete for a dwindling stock. BDPD⁰’s introduction cites this as empirical evidence for the temporal asymmetry of resource dynamics — alongside Perolat et al. (2017) and the Bardi Seneca framework — motivating the platform’s exploration of how agent composition and wealth inequality interact with asymmetric collapse. The BDPD Arena’s logistic engine and its P8 reactive-versus-conservative sweep directly test the mechanism Owusu et al. observe at the continuous aggregate level. The experiments use human subjects; they do not model the agent-level architecture heterogeneity that is BDPD’s central experimental variable.

2.12 Various

Dal Bó (2006). Regulatory capture: A review. Oxford Review of Economic Policy, 22(2), 203–225. DOI: 10.1093/oxrep/grj013

Regulatory capture: A review

This article reviews both the theoretical and empirical literatures on regulatory capture. The scope is broad, but utility regulation is emphasized. I begin by describing the Stigler–Peltzman approach to the economics of regulation. I then open the black box of influence and regulatory discretion using a three-tier hierarchical agency model under asymmetric information (in the spirit of Laffont and Tirole, 1993). I discuss alternative modelling approaches with a view to a richer set of positive predictions, including models of common agency, revolving doors, informational lobbying, coercive pressure, and influence over committees. I discuss empirical work involving capture and regulatory outcomes. I also review evidence on the revolving-door phenomenon and on the impact that different methods for selecting regulators appear to have on regulatory outcomes. The last section contains open questions for future research.

AI Note: Dal Bó reviews both theoretical and empirical literatures on regulatory capture — the process through which special interests, particularly regulated monopolies, manipulate the state agencies meant to control them. The survey covers the Stigler-Peltzman approach, Laffont-Tirole three-tier agency models under asymmetric information, common agency, revolving doors, and the impact of regulator selection methods on outcomes. BDPD⁰ maps this concept directly into its platform architecture: the wealth-weighted scheduler, in which richer agents act first, is explicitly designed to model regulatory capture — the structural mechanism by which accumulated wealth translates into institutional power that compounds inequality. The BDPD platform finding (P7) that scheduler has a negligible effect on the gate-pass rate (welfare p = 1.000) suggests that capture is secondary to agent-type composition in determining collapse. The review addresses capture in utility regulation and political economy, not in the context of common-pool resource governance with strategic extraction dynamics.

Fehr & Schmidt (1999). A theory of fairness, competition, and cooperation. The Quarterly Journal of Economics, 114(3), 817–868. DOI: 10.1162/003355399556151

A theory of fairness, competition, and cooperation

There is strong evidence that people exploit their bargaining power in competitive markets but not in bilateral bargaining situations. There is also strong evidence that people exploit free-riding opportunities in voluntary cooperation games. Yet, when they are given the opportunity to punish free riders, stable cooperation is maintained, although punishment is costly for those who punish. This paper asks whether there is a simple common principle that can explain this puzzling evidence. We show that if some people care about equity the puzzles can be resolved. It turns out that the economic environment determines whether the fair types or the selfish types dominate equilibrium behavior.

AI Note: Fehr and Schmidt develop a model of inequity aversion in which a fraction of the population derives disutility from both disadvantageous and advantageous inequity. The model shows that the economic environment — competitive markets versus bilateral bargaining versus public goods with punishment — determines whether fair types or selfish types dominate equilibrium behaviour. It provides a unified theoretical micro-foundation for the experimental regularities of altruistic punishment, wage rigidity, and cooperation in social dilemmas with a single preference parameter. BDPD⁰ cites the model to acknowledge that neither heuristic agents nor LLM agents capture the bounded rationality, emotional responses to visible inequality, or tacit social negotiation that Fehr-Schmidt formalises — a gap the Forest of Humbaba card game is designed to bridge. The model assumes a fixed distribution of fairness types within a human population; it does not address whether LLM agents exhibit inequity aversion, and if so, through what mechanism — training-data reflection of human norms versus emergent strategic reasoning.

Simon (1955). A behavioral model of rational choice. The Quarterly Journal of Economics, 69(1), 99–118. DOI: 10.2307/1884852

A behavioral model of rational choice

A model is proposed for the descritpion of rationa choice by organisms of limited computational ability

AI Note: Simon’s foundational paper introduces bounded rationality — the idea that decision-makers, constrained by limited cognitive capacity and information, satisfice rather than optimise. This reframing of rationality from substantive to procedural opened the research programme that became behavioural economics. BDPD⁰ and BDPD¹ cite Simon alongside Gigerenzer (2011) to frame the central dichotomy of the project: the contrast between fixed-policy heuristics (rule-bound) and signal-conditioned reasoning (adaptive). The BDPD platform’s experimental design tests which side of that divide outperforms on a fragile commons. The original paper addresses organisational decision-making in administrative contexts; it does not model the multi-agent strategic commons or LLM cognitive architectures.

Gigerenzer & Gaissmaier (2011). Heuristic decision making. Annual Review of Psychology, 62, 451–482. DOI: 10.1146/annurev-psych-120709-145346

Heuristic decision making

As reflected in the amount of controversy, few areas in psychology have undergone such dramatic conceptual changes in the past decade as the emerging science of heuristics. Heuristics are efficient cognitive processes, conscious or unconscious, that ignore part of the information. Because using heuristics saves effort, the classical view has been that heuristic decisions imply greater errors than do “rational” decisions as defined by logic or statistical models. However, for many decisions, the assumptions of rational models are not met, and it is an empirical rather than an a priori issue how well cognitive heuristics function in an uncertain world. To answer both the descriptive question (“Which heuristics do people use in which situations?”) and the prescriptive question (“When should people rely on a given heuristic rather than a complex strategy to make better judgments?”), formal models are indispensable. We review research that tests formal models of heuristic inference, including in business organizations, health care, and legal institutions. This research indicates that (a) individuals and organizations often rely on simple heuristics in an adaptive way, and (b) ignoring part of the information can lead to more accurate judgments than weighting and adding all information, for instance for low predictability and small samples. The big future challenge is to develop a systematic theory of the building blocks of heuristics as well as the core capacities and environmental structures these exploit.

AI Note: Gigerenzer and Gaissmaier review the theory and evidence for heuristic decision-making, arguing that simple heuristics can be ecologically rational — matching the structure of the decision environment — and can outperform more complex strategies under conditions of uncertainty. The review covers the accuracy-effort trade-off, the adaptive toolbox metaphor, and applications in medicine, law, and business. BDPD⁰ and BDPD¹ cite this work alongside Simon (1955) to frame the central dichotomy of the project: the contrast between fixed-policy heuristics and signal-conditioned reasoning. The BDPD platform’s experimental design tests which side of that divide — rule-bound or adaptive — outperforms the other on a fragile commons. The review synthesises evidence from human decision-making studies; it does not address whether LLM agents employ heuristics, learn new ones in-context, or rely on fundamentally different decision architectures.

Milinski et al. (2008). The collective-risk social dilemma and the prevention of simulated dangerous climate change. Proceedings of the National Academy of Sciences, 105(7), 2291–2294. DOI: 10.1073/pnas.0709546105

The collective-risk social dilemma and the prevention of simulated dangerous climate change

Will a group of people reach a collective target through individual contributions when everyone suffers individually if the target is missed? This “collective-risk social dilemma” exists in various social scenarios, the globally most challenging one being the prevention of dangerous climate change. Reaching the collective target requires individual sacrifice, with benefits to all but no guarantee that others will also contribute. It even seems tempting to contribute less and save money to induce others to contribute more, hence the dilemma and the risk of failure. Here, we introduce the collective-risk social dilemma and simulate it in a controlled experiment: Will a group of people reach a fixed target sum through successive monetary contributions, when they know they will lose all their remaining money with a certain probability if they fail to reach the target sum? We find that, under high risk of simulated dangerous climate change, half of the groups succeed in reaching the target sum, whereas the others only marginally fail. When the risk of loss is only as high as the necessary average investment or even lower, the groups generally fail to reach the target sum. We conclude that one possible strategy to relieve the collective-risk dilemma in high-risk situations is to convince people that failure to invest enough is very likely to cause grave financial loss to the individual. Our analysis describes the social window humankind has to prevent dangerous climate change.

AI Note: Milinski and colleagues introduce the collective-risk social dilemma, simulating climate-change prevention in a controlled experiment where groups must reach a fixed target sum through successive monetary contributions, facing a probability of losing all remaining money if the target is missed. They find that under high risk, about half of groups succeed; under low risk, groups generally fail. BDPD⁰‘s discussion cites this alongside Vasconcelos et al. (2014) to map the Tragedy of the Compensator onto real-world climate negotiations, where early movers’ conservation creates ecological slack that defectors exploit. BDPD¹’s D1 pilot contrasts the collective-risk setup against the BDPD cliff: in Milinski’s design, the threshold is collective and communication helps groups coordinate toward it; on the BDPD cliff, the threshold is individual — a single aggressor suffices — and communication cannot deter a player whose privately optimal strategy is over-extraction regardless of what others say.

Centola et al. (2018). Experimental evidence for tipping points in social convention. Science, 360(6393), 1116–1119. DOI: 10.1126/science.aas8827

Experimental evidence for tipping points in social convention

Once a population has converged on a consensus, how can a group with a minority viewpoint overturn it? Theoretical models have emphasized tipping points, whereby a sufficiently large minority can change the societal norm. Centola et al. devised a system to study this in controlled experiments. Groups of people who had achieved a consensus about the name of a person shown in a picture were individually exposed to a confederate who promoted a different name. The only incentive was to coordinate. When the number of confederates was roughly 25% of the group, the opinion of the majority could be tipped to that of the minority.

AI Note: Centola and colleagues provide direct empirical demonstration of tipping points in social conventions through an experimental system in which human subjects coordinate on naming a novel object. When a committed minority reached approximately 25% of the population, the new convention consistently spread to the entire network; below that threshold, the old convention persisted. The result confirms theoretical predictions that minority groups can overturn established social equilibria once they reach a critical mass. BDPD²’s V4 vignette cites this as the human-subject analogue for the active-contagion mechanism: a sufficiently visible minority can flip the majority’s behavioural equilibrium, paralleling the hypothesis that LLM conformists co-breach a pact after observing a defector’s visible extraction. The experiment uses a coordination game with fixed-payoff structure, not a commons dilemma with cumulative resource depletion and wealth feedback.

Boiko et al. (2023). Autonomous chemical research with large language models. Nature, 624(7992), 570–578. DOI: 10.1038/s41586-023-06792-0

Autonomous chemical research with large language models

Transformer-based large language models are making significant strides in various fields, such as natural language processing1,2,3,4,5, biology6,7, chemistry8,9,10 and computer programming11,12. Here, we show the development and capabilities of Coscientist, an artificial intelligence system driven by GPT-4 that autonomously designs, plans and performs complex experiments by incorporating large language models empowered by tools such as internet and documentation search, code execution and experimental automation. Coscientist showcases its potential for accelerating research across six diverse tasks, including the successful reaction optimization of palladium-catalysed cross-couplings, while exhibiting advanced capabilities for (semi-)autonomous experimental design and execution. Our findings demonstrate the versatility, efficacy and explainability of artificial intelligence systems like Coscientist in advancing research.

AI Note: Boiko and colleagues present Coscientist, a multi-LLM intelligent agent driven by GPT-4 that autonomously designs, plans, and executes complex chemistry experiments by integrating web search, code execution, and robotic lab automation. Across six tasks — including successful reaction optimisation of palladium-catalysed cross-couplings — Coscientist demonstrates that LLMs can function as autonomous scientific agents, not merely as conversational assistants. BDPD⁰’s methodological coda cites Coscientist as part of the evidence that frontier LLMs can act as substantive cognitive collaborators in original scientific research — a thesis that frames BDPD’s own LLM-orchestrated development process not as an anomaly but as an early instance of a broader shift in scientific practice. Coscientist operates in a well-defined, tool-constrained chemistry domain; its autonomy relies on structured APIs and documented protocols, not on the open-ended strategic social reasoning required in a commons governance setting.

Rahwan et al. (2019). Machine behaviour. Nature, 568(7753), 477–486. DOI: 10.1038/s41586-019-1138-y

Machine behaviour

Machines powered by artificial intelligence increasingly mediate our social, cultural, economic and political interactions. Understanding the behaviour of artificial intelligence systems is essential to our ability to control their actions, reap their benefits and minimize their harms. Here we argue that this necessitates a broad scientific research agenda to study machine behaviour that incorporates and expands upon the discipline of computer science and includes insights from across the sciences. We first outline a set of questions that are fundamental to this emerging field and then explore the technical, legal and institutional constraints on the study of machine behaviour.

AI Note: Rahwan and colleagues’ seminal Nature comment articulates the ‘machine behaviour’ programme: studying intelligent machines — including AI agents and algorithms — as empirical objects of behavioural inquiry, using the methods of behavioural science rather than treating them purely as engineered artefacts. BDPD¹’s introduction situates its LLM-agent experiments within this programme, treating DeepSeek-flash players not as task-solvers but as behavioural subjects whose cooperation, defection, and communication patterns can be systematically observed and decomposed. BDPD⁰’s methodological coda extends the programme to LLMs as scientific collaborators. The comment is a manifesto and research agenda; it does not provide experimental results or specific hypotheses about LLM behaviour in commons dilemmas.

Isaacs (1965). Differential Games: A Mathematical Theory with Applications to Warfare and Pursuit, Control and Optimization. John Wiley and Sons.

Differential Games: A Mathematical Theory with Applications to Warfare and Pursuit, Control and Optimization

AI Note: Isaacs’ foundational monograph develops the mathematical theory of differential games — strategic interactions in continuous time where players’ control variables affect the evolution of a dynamical system described by differential equations. The framework has been applied to pursuit-evasion, warfare, and economic competition. BDPD⁰‘s introduction cites Isaacs to clarify that the ’differential’ in BDPD’s name refers to the ODE-specified resource dynamics (logistic, Seneca), not to differential game theory in Isaacs’ sense. BDPD agents follow discrete-turn strategies rather than continuous-time optimal control, making the platform conceptually closer to agent-based modelling than to differential game theory. The book is a mathematical treatise; it does not model resource dilemmas, agent heterogeneity, or the institutional governance mechanisms central to BDPD.

DeepSeek-AI (2026). DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence.

DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence

Anthropic (2026). Claude Sonnet 4.6 Model Card.

Claude Sonnet 4.6 Model Card

Team (2026). Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving. URL: https://qwen.ai/blog?id=qwen3.6-max-preview

Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving

GLM-5-Team & others (2026). GLM-5: from Vibe Coding to Agentic Engineering. URL: https://arxiv.org/abs/2602.15763

GLM-5: from Vibe Coding to Agentic Engineering

Olson (1965). The Logic of Collective Action: Public Goods and the Theory of Groups. Harvard University Press.

The Logic of Collective Action: Public Goods and the Theory of Groups

AI Note: Olson’s landmark book argues that rational, self-interested individuals will not act to achieve common group interests unless the group is small, or coercion or selective incentives are present. His tripartite taxonomy — privileged, intermediate, and latent groups — turns on the noticeability of individual contributions, not cardinal group size. BDPD⁰‘s introduction uses Olson to anchor the proposition that inequality may facilitate collective action (the ’Olson effect’), with a wealthy actor bearing provision costs while others free-ride. BDPD’s mini-course Lesson 2 — “We Are Enough Among Ourselves” — critiques the strong reading of Olson, using the experimental record (Balliet 2010, Marwell-Ames 1981) and Dunbar’s number to argue that the constraint is informational and cognitive, not numerical. The argument addresses human groups with fixed cognitive architectures; the BDPD platform tests whether LLM agents with different perceptual architectures shift the privileged-intermediate-latent boundary.

Dunbar (1993). Coevolution of neocortical size, group size and language in humans. Behavioral and Brain Sciences, 16(4), 681–735. DOI: 10.1017/S0140525X00032325

Coevolution of neocortical size, group size and language in humans

AI Note: Dunbar demonstrates that group size in primates is a function of relative neocortical volume, extrapolating to a predicted human group size of approximately 150 — ‘Dunbar’s number’ — argued to reflect a cognitive ceiling on the number of stable social relationships an individual can maintain. He proposes that language evolved as a more efficient bonding mechanism than physical grooming. BDPD’s mini-course Lesson 2 — “We Are Enough Among Ourselves” — uses Dunbar’s number to reframe Olson’s group-size argument: the constraint on cooperation is not cardinal group size but the cognitive substrate that tracks interaction histories, detects defection, and sustains reputation-based strategies. For LLM agents in BDPD, this cognitive ceiling may operate differently — LLMs have no neocortical limit but face prompt-window constraints, attention decay, and training-distribution biases. The paper addresses primate social cognition, not artificial-agent governance.

Marwell & Ames (1981). Economists free ride, does anyone else?: Experiments on the provision of public goods, IV. Journal of Public Economics, 15(3), 295–310. DOI: 10.1016/0047-2727(81)90013-X

Economists free ride, does anyone else?: Experiments on the provision of public goods, IV

AI Note: Marwell and Ames conduct one of the earliest large-scale public goods experiments, testing voluntary contributions in groups of up to 80 subjects — far beyond the ‘quite small’ groups Olson predicted could self-organise. They find measurable cooperation at group sizes that Olson’s logic-of-collective-action framework would classify as latent and incapable of self-provision. BDPD’s mini-course Lesson 2 — “We Are Enough Among Ourselves” — uses this finding as empirical evidence against the strong reading of Olson: if 80 subjects can sustain cooperation, the constraint is not cardinal group size but something else — visibility of contributions, monitoring capacity, or the cognitive architecture of the agents. The experiment uses human subjects with a step-level public good; it does not model the resource dynamics central to the BDPD commons.

Henrich et al. (2006). Costly Punishment Across Human Societies. Science, 312, 1767–1770.

Costly Punishment Across Human Societies

AI Note: Henrich and colleagues run public-goods-with-punishment experiments across fifteen diverse small-scale societies, finding that (i) all populations demonstrate some willingness to engage in costly punishment as unequal behaviour increases, (ii) punishment magnitude varies substantially across populations, and (iii) costly punishment positively covaries with altruistic behaviour. BDPD’s mini-course Lesson 4 — “Let’s Punish the Cheaters” — cites this as the most ambitious replication of the Fehr-Gächter result, demonstrating that costly punishment is a human universal. However, the cross-cultural variation also bounds the altruistic-punishment intuition: punishment can be deployed against cooperators, and its effect depends on who wields it and within what normative framework. The experiments use one-shot public goods with human subjects; they do not address whether LLM agents from different training distributions exhibit analogous cross-model variation in punishment behaviour.

Herrmann et al. (2008). Antisocial Punishment Across Societies. Science, 319(5868), 1362–1367. DOI: 10.1126/science.1153808

Antisocial Punishment Across Societies

AI Note: Herrmann and colleagues document antisocial punishment across sixteen subject pools worldwide: in some societies, cooperators are punished by free-riders — the opposite of the altruistic-punishment pattern. Antisocial punishment correlates negatively with norms of civic cooperation and rule of law at the country level, and its presence weakens the cooperation-enhancing effect of punishment. BDPD’s mini-course Lesson 4 cites this alongside Henrich et al. (2006) to bound the punishment intuition from the opposite direction: punishment is not inherently pro-social, and its effect depends on the normative framework within which it is deployed. BDPD¹’s graduated-sanction pilots avoid this by embedding punishment in an institutional (pact-based) rather than peer-to-peer framework. The study uses human subjects in one-shot games; it does not model punishment dynamics under the cumulative resource extraction and irreversible collapse that characterise the BDPD commons.

Black (1948). On the Rationale of Group Decision-making. Journal of Political Economy, 56(1), 23–34. DOI: 10.1086/256633

On the Rationale of Group Decision-making

AI Note: Black’s foundational paper establishes that under single-peaked preferences on a single dimension, majority rule produces a stable, transitive outcome: the preference of the median voter. This median-voter theorem is the most important positive result in voting theory, identifying the precise condition — single-peakedness on one dimension — under which the intuitively simplest voting rule works. BDPD’s mini-course Lesson 5 — “Democratic Voting” — juxtaposes Black’s theorem against Arrow’s impossibility and McKelvey’s chaos results to illustrate that the binding variable is not the voting rule but the dimensionality of the preference space. For LLM-agents in BDPD, the lesson argues, dimensionality is determined by prompt structure — an engineering choice rather than a fixed feature of an electorate, with direct implications for how governance simulations are designed. The original paper addresses human committees; the generalisation to LLM voters is the BDPD extrapolation, not Black’s claim.

Sen (1970). The Impossibility of a Paretian Liberal. Journal of Political Economy, 78(1), 152–157. DOI: 10.1086/259614

The Impossibility of a Paretian Liberal

AI Note: Sen proves the impossibility of a Paretian liberal: no social decision function can simultaneously respect individual rights (each person being decisive over at least one pair of alternatives) and the weak Pareto principle (if everyone prefers x to y, society should prefer x to y). BDPD’s mini-course Lesson 5 — “Democratic Voting” — uses this result to complete the impossibility triptych (Arrow, Gibbard-Satterthwaite, Sen) that bounds the good-will intuition about democratic voting. For LLM-based governance in BDPD, the Sen paradox raises a substantive question: if individual agents have ‘rights’ over their harvest decisions, can the collective outcome respect both Pareto efficiency and individual decisiveness? The theorem is formal social choice theory; it does not model resource dynamics.

Sen (1970). Collective Choice and Social Welfare. Holden-Day.

Collective Choice and Social Welfare

AI Note: Sen’s treatise is the standard formalisation of Arrow’s social-choice framework and the source of the workhorse restatement of Arrow’s theorem used in BDPD’s mini-course Lesson 5 — “Democratic Voting”. A social welfare function is required to map every admissible profile of individual orderings to a complete, transitive social ordering (what Sen calls “collective rationality”); Arrow’s impossibility theorem then shows that no such function can simultaneously satisfy four conditions — Unrestricted Domain (U), the Pareto Principle (P), Independence of Irrelevant Alternatives (I), and Non-dictatorship (D). Sen states the result with these four axioms (“these mild-looking axioms U, I, P and D cannot be simultaneously fulfilled”); transitivity is part of the definition of a social ordering, not a separate fifth axiom — the distinction Lesson 5 follows. This work is distinct from Sen’s “liberal paradox” (the impossibility of a Paretian liberal), which BDPD cites separately as sen1970impossibility. The original is Holden-Day, San Francisco, 1970; an expanded edition appeared in 2017. The book is formal social choice theory and does not model resource dynamics; its application to LLM voters in BDPD is an extrapolation, not Sen’s claim.

Plott (1967). A Notion of Equilibrium and Its Possibility Under Majority Rule. American Economic Review, 57(4).

A Notion of Equilibrium and Its Possibility Under Majority Rule

AI Note: Plott proves that under majority rule with multi-dimensional policy spaces, an equilibrium exists only under a highly restrictive symmetry condition on preference distributions — the Plott symmetry condition — which almost never holds in practice. BDPD’s mini-course Lesson 5 — “Democratic Voting” — uses this result as the negative counterpart to Black’s positive median-voter theorem: on one dimension, majority rule works; on multiple dimensions, it requires a symmetry that is empirically absent. The Fiorina-Plott experimental tests confirmed that outcomes cluster near the equilibrium only under conditions approximating this symmetry. The theorem addresses formal voting with fixed preferences; it does not address how LLM agents with prompt-structured preference orderings behave under majority rule.

Gibbard (1973). Manipulation of Voting Schemes: A General Result. Econometrica, 41, 587–601.

Manipulation of Voting Schemes: A General Result

AI Note: Gibbard proves that any non-dictatorial voting rule with at least three alternatives is manipulable: there exist preference profiles under which a voter can obtain a better outcome by misrepresenting their preferences. This Gibbard-Satterthwaite theorem is the strategic counterpart to Arrow’s impossibility — not only can no rule aggregate preferences perfectly, but no rule can prevent voters from gaming whatever rule is in place. BDPD’s mini-course Lesson 5 — “Democratic Voting” — uses this theorem to raise the question of strategic sophistication: LLM voters can compute strategic misrepresentations in milliseconds, while human voters typically cannot. The prediction is that LLM committees converge to strategic equilibria that human committees never reach — a hypothesis the BDPD platform is positioned to test. The theorem addresses formal voting rules under complete information; it assumes voters know others’ preferences and the decision rule, assumptions that may not hold in the BDPD commons setting where agents observe extraction outcomes, not preferences.

Satterthwaite (1975). Strategy-proofness and Arrow’s conditions: Existence and correspondence theorems for voting procedures and social welfare functions. Journal of Economic Theory, 10(2), 187–217. DOI: 10.1016/0022-0531(75)90050-2

Strategy-proofness and Arrow’s conditions: Existence and correspondence theorems for voting procedures and social welfare functions

AI Note: Satterthwaite proves, independently of Gibbard (1973), that any non-dictatorial voting rule with at least three alternatives is strategically manipulable: there exist preference profiles where a voter can obtain a better outcome by misrepresenting their preferences. The Gibbard-Satterthwaite theorem is the voting-theory counterpart to Arrow’s impossibility. BDPD’s mini-course Lesson 5 — “Democratic Voting” — uses this result to raise the question of strategic sophistication among LLM voters: agents that can compute strategic misrepresentations in milliseconds may converge to equilibria human committees never reach. The theorem addresses formal voting rules under complete information; it does not consider how agenda-setting interacts with strategic voting in multi-agent resource settings.

McKelvey (1976). Intransitivities in multidimensional voting models and some implications for agenda control. Journal of Economic Theory, 12(3), 472–482. DOI: 10.1016/0022-0531(76)90040-5

Intransitivities in multidimensional voting models and some implications for agenda control

AI Note: McKelvey proves the chaos theorem for multi-dimensional voting: under majority rule with three or more alternatives and preferences that are not single-peaked with Plott symmetry, any alternative can be reached from any other through a sequence of majority-rule defeats. The outcome depends entirely on agenda control, and a skilled agenda-setter can engineer any result. BDPD’s mini-course Lesson 5 — “Democratic Voting” — uses McKelvey alongside Black (1948) to demonstrate that the binding variable is not the voting rule but the dimensionality of the preference space. For LLM voters in BDPD, dimensionality is determined by prompt structure — making it an engineering choice. The theorem addresses formal voting with fixed preferences; it does not consider how preferences themselves evolve under strategic interaction or resource feedback.

Fiorina & Plott (1978). Committee Decisions under Majority Rule: An Experimental Study. American Political Science Review, 72, 575–598.

Committee Decisions under Majority Rule: An Experimental Study

AI Note: Fiorina and Plott report experiments on committee decision-making under majority rule with fixed member preferences. Their headline finding is that the core — the majority-rule equilibrium — consistently performs best among predictive models, but a subtler result matters more: even when the theoretical equilibrium does not exist, outcomes remain bounded within small regions of the policy space rather than exhibiting the chaos that McKelvey’s theorem would suggest. BDPD’s mini-course Lesson 5 — “Democratic Voting” — uses this study to argue that the theoretical chaos of multi-dimensional voting is empirically bounded by features the theory does not model, opening the question of what bounds it — and whether LLM voters respect the same bounds. The experiments use small five-member committees with human subjects and induced preferences; the generalisation to LLM agents with prompt-structured preference orderings remains untested.

Goodhart (1984). Problems of Monetary Management: The UK Experience. Macmillan Education UK. DOI: 10.1007/978-1-349-17295-5_4

Problems of Monetary Management: The UK Experience

AI Note: Goodhart’s chapter, based on his experience as a Bank of England economist during the UK’s 1970s monetary targeting episode, articulates the mechanism underlying Goodhart’s Law: when a statistical regularity (the M3-inflation relationship) is made a policy target, the behavioural patterns of financial institutions adapt to the new incentive structure, and the regularity dissolves. Unlike Campbell’s Law, which operates through measurement corruption, Goodhart’s mechanism works through the restructuring of activity: agents change what they do, not how it is reported. BDPD’s mini-course Lesson 7 — “We Need a Rule” — uses Goodhart’s Law as the foundational argument that rule-bound governance is systematically vulnerable when agents can adapt to the rule: the BDPD platform’s finding that graduated sanctions outperform flat ones is a governance analogue, where the sanction schedule must anticipate strategic adaptation. The paper addresses monetary policy in a specific institutional context; generalising to commons governance with heterogeneous adaptive agents is the BDPD extrapolation.

Strathern (1997). `Improving ratings’: audit in the British University system. European Review, 5(3), 305–321. DOI: 10.1002/(SICI)1234-981X(199707)5:3<305::AID-EURO184>3.0.CO;2-4

`Improving ratings’: audit in the British University system

AI Note: Strathern’s anthropological essay on audit in British universities documents how performance metrics transform the institutions they measure: universities restructured hiring, teaching, and administration to optimise on research-assessment scores and student-satisfaction ratings. She coined the now-standard formulation of Goodhart’s Law — “When a measure becomes a target, it ceases to be a good measure” — naming the phenomenon as a cultural dynamic, not merely a statistical one. BDPD’s mini-course Lesson 7 — “We Need a Rule” — uses Strathern alongside Goodhart, Campbell, and Lucas to complete the four-pillar argument against rule-bound governance. The BDPD platform’s graduated-sanction result — escalation shape matters more than flat severity — can be read as a governance design that anticipates the audit-culture dynamic. The essay addresses higher education; it does not model commons governance.

Campbell (1979). Assessing the impact of planned social change. Evaluation and Program Planning, 2(1), 67–90. DOI: 10.1016/0149-7189(79)90048-X

Assessing the impact of planned social change

AI Note: Campbell’s classic paper on social-programme evaluation articulates what became known as Campbell’s Law: the more any quantitative social indicator is used for social decision-making, the more it will be subject to corruption pressures and distort the social processes it is intended to monitor. Unlike Goodhart’s Law — where agents restructure activities to move the measured quantity — Campbell’s mechanism targets the measurement process itself: agents falsify reports, teach to the test, or reclassify offences at the point of data collection. BDPD’s mini-course Lesson 7 — “We Need a Rule” — juxtaposes Campbell’s Law with Goodhart’s Law to frame the central tension: rule-bound governance, whether via fixed sanctions or fixed monitoring targets, creates incentive structures that degrade the very signals the rules depend on. The paper addresses social indicators in public administration, not the dynamic resource signals of a Seneca-governed commons; the extension to leading/lagging indicator selection in complex adaptive systems is a BDPD extrapolation.

Holmstrom (1979). Moral Hazard and Observability. Bell Journal of Economics, 10(1), 74–91.

Moral Hazard and Observability

AI Note: Holmström’s classic paper formalises the single-task moral-hazard problem under imperfect observability, establishing the theoretical foundation for principal-agent models: when the principal observes only a noisy signal of the agent’s effort, the optimal contract trades off insurance against incentives (the informativeness principle). It is the foundation on which the later multitask extension (see holmstrom1991multitask) builds, and is cited in the mini-course for the principal-agent lineage rather than for the multitasking result itself. The model addresses contractual relationships, not the resource-coupled strategic interactions of the BDPD commons.

Holmstrom & Milgrom (1991). Multitask Principal-Agent Analyses: Incentive Contracts, Asset Ownership, and Job Design. Journal of Law, Economics, & Organization, 7, 24–52.

Multitask Principal-Agent Analyses: Incentive Contracts, Asset Ownership, and Job Design

AI Note: Holmström and Milgrom extend the single-task principal-agent model (see holmstrom1979moral) to the case where the agent allocates effort across multiple tasks and the principal can observe and reward only a subset. The central result is that when efforts are substitutes in the agent’s cost function, providing strong incentives on a measurable task drains effort from unmeasured ones — so the optimal contract may deliberately weaken incentives on the measurable dimension. BDPD’s mini-course Lesson 7 — “We Need a Rule” — uses this multitask model to formalise the Goodhart mechanism: a target on one observable dimension degrades the unobserved processes the agent also controls, and the distortion grows with the stakes attached to the measured dimension and with the agent’s computational capacity to identify and exploit the unmonitored channels. In BDPD terms, a flat sanction on harvest violations creates a single observable dimension, while a graduated ladder may preserve multiple behavioural channels that a flat target collapses. The model addresses contractual relationships, not the resource-coupled strategic interactions of the BDPD commons.

Lucas (1976). Econometric policy evaluation: A critique. Carnegie-Rochester Conference Series on Public Policy, 1, 19–46. DOI: 10.1016/S0167-2231(76)80003-6

Econometric policy evaluation: A critique

AI Note: Lucas argues that econometric models used for policy evaluation contain reduced-form parameters — not structural constants — that reflect agents’ optimal behaviour under the existing policy regime. When the regime changes, agents re-optimise, and the reduced-form parameters shift, invalidating forecasts based on historical relationships. BDPD’s mini-course Lesson 7 — “We Need a Rule” — positions the Lucas critique alongside Goodhart’s Law and Campbell’s Law as the third pillar of the argument against rule-bound governance: if agents anticipate the rule, they adapt their behaviour to it, and the statistical relationship the rule was built on dissolves. The critique addresses macroeconomic policy evaluation; its extension to commons governance with heterogeneous adaptive agents is a BDPD extrapolation. The original argument assumes rational expectations and complete regime knowledge.

Manheim & Garrabrant (2019). Categorizing Variants of Goodhart’s Law. URL: https://arxiv.org/abs/1803.04585

Categorizing Variants of Goodhart’s Law

AI Note: Manheim and Garrabrant categorise four distinct mechanisms of Goodhart’s Law — regressional, extremal, causal, and adversarial — distinguishing failure modes that are often conflated under the single label. Each mechanism implies a different class of governance failure and a different class of remedies. BDPD’s mini-course Lesson 7 — “We Need a Rule” — uses this taxonomy as the most systematic modern classification of Goodhart effects, connecting monetary policy (Goodhart), social indicators (Campbell), econometric policy (Lucas), and AI alignment (Manheim-Garrabrant) into a unified framework. The BDPD platform’s graduated-sanction result — escalation shape beats flat amount — can be reframed through this lens as a governance design that anticipates adversarial Goodhart effects. The taxonomy is conceptual; it does not provide empirical tests of which mechanism dominates in specific governance settings.

Pigou (1920). The Economics of Welfare. Macmillan.

The Economics of Welfare

AI Note: Pigou’s landmark treatise introduces the distinction between private and social net product, establishing the economic rationale for externality-correcting taxation: when a factory’s emissions impose costs on neighbours that the factory owner does not bear, a tax calibrated to the marginal social damage restores efficient resource allocation. BDPD’s mini-course Lesson 6 — “The Market Knows” — uses Pigou as the first pillar of the Pigou-Coase-Demsetz triptych, arguing that all three solutions presuppose a particular kind of agent: the Pigovian regulator needs omniscience about damage curves. BDPD³’s finding that the instrument (cap, levy, fine) matters less than the trigger signal repositions Pigou’s question — the binding constraint is not calibrating the tax but monitoring the right variable. The treatise addresses welfare economics at the societal level; it does not model agent-level strategic responses.

Coase (1960). The Problem of Social Cost. The Journal of Law and Economics, III, 1-44.

The Problem of Social Cost

This paper is concerned with those actions of business firms which have harmful effects on others. The standard example is that of a factory the smoke from which has harmful effects on those occupying neighbouring properties. The economic analysis of such a situation has usually proceeded in terms of a divergence between the private and social product of the factory, in which economists have largely followed the treatment of Pigou in The Economics of Welfare. The conclusions to which this kind of analysis seems to have led most economists is that it would be desirable to make the owner of the factory liable for the damage caused to those injured by the smoke, or alternatively, to place a tax on the factory owner varying with the amount of smoke produced and equivalent in money terms to the damage it would cause, or finally, to exclude the factory from residential districts (and presumably from other areas in which the emission of smoke would have harmful effects on others). It is my contention that the suggested courses of action are inappropriate, in that they lead to results which are not necessarily, or even usually, desirable.

AI Note: Coase’s landmark paper reframes the problem of externalities as reciprocal: the question is not who causes harm, but which resource use is more valuable. He demonstrates that if property rights are well-defined and transaction costs are zero, parties will bargain to the efficient outcome regardless of the initial allocation of rights — the Coase Theorem. This overturned the Pigovian presumption that externalities require state taxation or regulation. BDPD’s mini-course Lesson 6 — “The Market Knows” — uses Coase to build the thesis that both Pigovian and Coasean solutions presuppose a particular kind of agent: the Pigovian regulator needs omniscience about damage curves; the Coasean bargainers need low transaction costs and the cognitive capacity to negotiate. The BDPD platform’s finding that LLM agents do not spontaneously bargain to efficient outcomes even with a cheap-talk channel tests the Coasean premise directly. The paper addresses legal and economic theory, not computational agents in resource dilemmas.

Demsetz (1967). Toward a Theory of Property Rights. American Economic Review, 57(2), 347–359.

Toward a Theory of Property Rights

AI Note: Demsetz develops an economic theory of property rights, arguing that they emerge when the gains from internalising externalities exceed the costs of defining and enforcing the rights. Property rights are not simply assigned by a regulator but evolve as a society’s cost-benefit calculus shifts — a dynamic extension of the Coasean logic. BDPD’s mini-course Lesson 6 — “The Market Knows” — uses Demsetz to complete the Pigou-Coase-Demsetz triptych: if property rights emerge endogenously when it is efficient for them to do so, then the commons tragedy might resolve itself without institutional intervention, provided the gains from internalisation are large enough. The BDPD platform’s finding that a single aggressive agent dooms the commons regardless of initial endowments tests this premise: even well-defined de facto property rights (accumulated wealth advantage) fail to prevent collapse when enforcement is absent. The paper addresses property-rights theory in law and economics, not the dynamics of renewable-resource commons.

Stavins (1998). What Can We Learn from the Grand Policy Experiment? Lessons from SO2 Allowance Trading. Journal of Economic Perspectives, 12(3), 69–88. DOI: 10.1257/jep.12.3.69

What Can We Learn from the Grand Policy Experiment? Lessons from SO2 Allowance Trading

AI Note: Stavins examines the US SO₂ allowance trading programme — the first large-scale application of market-based environmental instruments — extracting lessons about the performance of cap-and-trade relative to command-and-control regulation. The programme achieved substantial cost savings and demonstrated that market-based instruments can work in practice. BDPD’s mini-course Lesson 6 — “The Market Knows” — cites Stavins to anchor the informational objection to Pigovian taxation: the regulator must know the marginal damage and abatement cost curves, information that is rarely available, making quantity instruments (cap-and-trade) more practical than price instruments (taxes). The paper addresses real-world environmental policy; it does not model agent-level strategic responses to market signals or the leading/lagging distinction central to BDPD³.

Farrell & Rabin (1996). Cheap Talk. Journal of Economic Perspectives, 10(3), 103–118. DOI: 10.1257/jep.10.3.103

Cheap Talk

AI Note: Farrell and Rabin’s canonical JEP article surveys the theory of cheap talk — costless, non-binding, pre-play communication — demonstrating that talk can be informative and Pareto-improving when players’ interests are sufficiently aligned, but only in games with multiple rankable equilibria where communication functions as a coordination device. Their central insight is that cheap talk expands the equilibrium set, selecting the Pareto-superior equilibrium when one exists, but has no positive prediction when the game lacks the right structural preconditions. BDPD¹’s D1 pilot operationalises this: the BDPD cliff configuration has no alternative equilibrium for talk to select, so Farrell-Rabin predict a null — and BDPD¹ empirically delivers one (d = 0.17 for LLM chat vs the classical d = 0.46 for written communication among humans). The resulting decomposition isolates agent architecture as the dominant cooperation driver. The paper addresses theory, not experiments; it predates and does not consider LLM agents.

Crawford (1998). A Survey of Experiments on Communication via Cheap Talk. Journal of Economic Theory, 78(2), 286–298. DOI: 10.1006/jeth.1997.2359

A Survey of Experiments on Communication via Cheap Talk

AI Note: Crawford surveys experimental evidence on behaviour in games with cheap-talk communication, focusing on environments where messages have no direct payoff implications. The survey covers sender-receiver games, coordination games, and bargaining experiments, establishing that cheap talk can be informative when players’ preferences are not too far apart and can serve as a coordination device in games with multiple Pareto-ranked equilibria. BDPD¹’s introduction uses Crawford alongside Farrell-Rabin (1996) to articulate the structural precondition for cheap-talk effectiveness: multiple rankable equilibria. The BDPD cliff configuration lacks this precondition — the unique sustainable equilibrium sits below the privately optimal harvest — so theory predicts a cheap-talk null, which the D1 pilot confirms (d = 0.17). The survey covers experimental evidence through the mid-1990s with human subjects exclusively; it predates both the modern LLM-agent literature and the specific fragile-commons setting that BDPD tests.

2.13 Cascading Tipping Points and Coupled Tipping Elements

Wunderling et al. (2024). Climate tipping point interactions and cascades: a review. Earth System Dynamics, 15, 41–74. DOI: 10.5194/esd-15-41-2024

Climate tipping point interactions and cascades: a review

AI Note: Wunderling and colleagues synthesise the growing literature on tipping cascades in coupled environmental systems, identifying the key challenge as understanding propagation — how one subsystem crossing a threshold destabilises others linked to it. BDPD³’s introduction and the mini-course Lessons 3 and 10 cite this review alongside Lenton et al. (2019) and Armstrong McKay et al. (2022) to establish that the risk of greatest concern is propagation, not isolated tipping — precisely the problem BDPD³’s leading-indicator governance addresses. The review is a literature synthesis; it catalogues interactions among tipping elements but does not evaluate governance interventions or model agent behaviour.

Armstrong McKay et al. (2022). Exceeding 1.5°C global warming could trigger multiple climate tipping points. Science, 377(6611), eabn7950. DOI: 10.1126/science.abn7950

Exceeding 1.5°C global warming could trigger multiple climate tipping points

AI Note: Armstrong McKay and colleagues synthesise paleoclimate, observational, and model-based evidence to reassess climate tipping-point thresholds, providing a revised shortlist of global ‘core’ tipping elements and regional ‘impact’ tipping elements. They find that current warming of ~1.1°C already lies within the lower uncertainty ranges of some tipping points, with multiple elements at risk in the Paris 1.5–2°C range. BDPD³’s introduction cites this work as part of the cascading-tipping literature that establishes the propagation problem: the risk is not isolated collapse of a single element, but the destabilisation of coupled subsystems. Lesson 10 of the mini-course draws on the systematic catalogue of interactions to motivate the leading-versus-lagging signal distinction. The paper catalogues tipping-element interactions but does not model or evaluate the governance interventions that could prevent propagation — the gap BDPD³’s leading-indicator result addresses.

Lenton et al. (2019). Climate tipping points — too risky to bet against. Nature, 575, 592–595. DOI: 10.1038/d41586-019-03595-0

Climate tipping points — too risky to bet against

AI Note: Lenton and colleagues’ high-profile Nature comment argues that climate tipping points — the loss of the West Antarctic ice sheet, Amazon dieback, Atlantic thermohaline collapse — are interconnected and could trigger each other in domino-like sequences. The framing is urgent: the consideration of tipping points ‘helps to define that we are in a climate emergency.’ BDPD³ and the mini-course cite this as the landmark statement that shifted the cascading-tipping literature from studying isolated elements to studying propagation. BDPD³’s introduction uses it to establish that the risk of greatest concern is no longer isolated collapse but the destabilisation of coupled subsystems — precisely the problem its leading-indicator governance addresses. The comment is a synthesis and call to action, not a formal model.

Wunderling et al. (2021). Interacting tipping elements increase risk of climate domino effects under global warming. Earth System Dynamics, 12, 601–619. DOI: 10.5194/esd-12-601-2021

Interacting tipping elements increase risk of climate domino effects under global warming

AI Note: Wunderling and colleagues investigate interacting tipping elements in a stylised four-element climate network, finding that coupling tends to destabilise the elements, making cascading transitions more likely than isolated tipping. BDPD³’s model section cites this alongside Klose et al. (2019) as a direct precedent for the finding that interacting tipping elements destabilise each other — the phenomenon the BDPD A→B/C configuration reproduces and on which it then tests governance interventions. The study uses a conceptual climate model without strategic agents or regulatory mechanisms; the BDPD contribution is adding the governance layer — what signal should the regulator monitor? — to the cascade substrate.

Klose et al. (2020). Emergence of cascading dynamics in interacting tipping elements of ecology and climate. Royal Society Open Science, 7, 200599. DOI: 10.1098/rsos.200599

Emergence of cascading dynamics in interacting tipping elements of ecology and climate

AI Note: Klose and colleagues investigate the emergence of cascading dynamics in interacting tipping elements using a conceptual lake-chain model as an ecological analogue. They demonstrate how tipping in one element can trigger domino-like transitions across coupled subsystems, characterising the conditions under which cascades emerge. BDPD³ cites this work as a direct ecological analogue of the A→B/C pollution-cascade setup: the lake-chain model’s unidirectional coupling maps onto the emitter-to-downwind topology that BDPD³ uses to test leading-versus-lagging governance. The model uses deterministic differential equations without strategic agents or governance interventions; BDPD³ adds both the agent layer and the regulator layer to the cascade substrate.

Klose et al. (2021). What do we mean, `tipping cascade’?. Environmental Research Letters, 16(12), 125002. DOI: 10.1088/1748-9326/ac3955

What do we mean, `tipping cascade’?

AI Note: Klose and colleagues formalise the dynamical anatomy of tipping cascades, distinguishing two-phase, domino, and joint cascade patterns according to how a critical transition propagates from one element to the next. In a two-phase cascade, the leader collapses first and the follower collapses only after a delay — the delay being the structural feature that makes governance possible if the signal is read early enough. BDPD³ uses this taxonomy across its model, results, and robustness sections: the A→B/C configuration is explicitly a two-phase cascade, and the rate-induced boundary at dt ≥ 3.5 occupies a gap in the taxonomy (which excludes rate-induced effects). The paper provides the conceptual vocabulary — leading, following, two-phase cascade — that structures BDPD³’s governance question. The taxonomy addresses Earth-system tipping elements; it does not model governance interventions or agent behaviour.

Wunderling et al. (2023). Global warming overshoots increase risks of climate tipping cascades in a network model. Nature Climate Change, 13, 75–82. DOI: 10.1038/s41558-022-01545-9

Global warming overshoots increase risks of climate tipping cascades in a network model

Wunderling et al. (2020). Basin stability and limit cycles in a conceptual model for climate tipping cascades. New Journal of Physics, 22, 123031. DOI: 10.1088/1367-2630/abc98a

Basin stability and limit cycles in a conceptual model for climate tipping cascades

Krönke et al. (2020). Dynamics of tipping cascades on complex networks. Physical Review E, 101, 042311. DOI: 10.1103/PhysRevE.101.042311

Dynamics of tipping cascades on complex networks

AI Note: Krönke and colleagues investigate how network topology affects the vulnerability of coupled multistable subsystems to tipping cascades. Using Erdős-Rényi, Watts-Strogatz, and Barabási-Albert networks — plus a moisture-recycling network of the Amazon — they find that clustering and spatial organisation increase vulnerability, and that even simple network structures can produce cascading tipping when coupling exceeds a critical strength. BDPD³’s model section cites this to justify its minimal three-arena directed graph: the simplest topology that admits a leading-following cascade with two independent victims. The study models passive tipping-element networks without strategic agents or governance; BDPD³ adds agent extraction and regulatory intervention to the cascade substrate.

Sinet et al. (2025). Approximating the bifurcation diagram of weakly and strongly coupled leading-following systems. Chaos, 35(6), 063135. DOI: 10.1063/5.0269773

Approximating the bifurcation diagram of weakly and strongly coupled leading-following systems

AI Note: Sinet and colleagues characterise the leading-following structure of unidirectionally coupled subsystems, providing a formal framework for systems where one component drives another. BDPD³’s model section uses this framework to classify the A→B/C pollution-cascade topology: A is the leading subsystem (emitter), B and C are following subsystems (downwind), coupled unidirectionally via pollution import. The leading-following vocabulary — combined with Klose et al.’s two-phase cascade taxonomy — provides the structural underpinning for BDPD³’s central governance question: which signal inside the leading subsystem should the regulator monitor? The framework addresses dynamical systems theory; it does not model governance interventions or strategic agents.

Lohmann et al. (2021). Abrupt climate change as a rate-dependent cascading tipping point. Earth System Dynamics, 12, 819–835. DOI: 10.5194/esd-12-819-2021

Abrupt climate change as a rate-dependent cascading tipping point

AI Note: Lohmann and colleagues propose a conceptual model of abrupt climate change as a rate-dependent cascading tipping point, distinguishing bifurcation-induced tipping (crossing a static threshold) from rate-induced tipping (the rate of change exceeds the system’s tracking capacity). Their sea ice-ocean circulation model shows that rate-induced tipping can occur before any bifurcation point is reached, with implications for early-warning detection. BDPD³’s robustness section and the mini-course Lesson 10 use this distinction to frame the governance-rate boundary: the win-win regime degrades at dt ≥ 3.5 not because a static threshold is crossed, but because the regulator’s sampling rate loses pace with the capital dynamics. The model studies climate subsystems without strategic agents or governance; BDPD³ extends the rate-induced concept to the governance–monitoring rate axis.

Klose et al. (2024). Rate-induced tipping cascades arising from interactions between the Greenland Ice Sheet and the Atlantic Meridional Overturning Circulation. Earth System Dynamics, 15, 635–652. DOI: 10.5194/esd-15-635-2024

Rate-induced tipping cascades arising from interactions between the Greenland Ice Sheet and the Atlantic Meridional Overturning Circulation

Lenton et al. (2024). Remotely sensing potential climate change tipping points across scales. Nature Communications, 15, 828. DOI: 10.1038/s41467-023-44609-w

Remotely sensing potential climate change tipping points across scales

2.14 Polycentric Governance (Modern)

Ostrom (2010). Beyond Markets and States: Polycentric Governance of Complex Economic Systems. American Economic Review, 100(3), 641–672. DOI: 10.1257/aer.100.3.641

Beyond Markets and States: Polycentric Governance of Complex Economic Systems

AI Note: Ostrom’s Nobel lecture, published as ‘Beyond Markets and States,’ articulates the polycentric governance vision in its mature form: multiple centres of decision-making, each with some autonomy, interacting within an overarching set of rules. Drawing on a half-century of research — from Vincent Ostrom’s 1961 metropolitan governance work to her own commons studies — she argues that polycentric systems outperform both pure centralisation and pure decentralisation because they exploit local knowledge while retaining coordination capacity. BDPD² uses this as the Bloomington-school definition of polycentricity anchoring its nested-World architecture. BDPD³’s model section follows Ostrom’s observation that polycentric systems succeed or fail not only on which rules they deploy but on how they monitor the variables those rules act upon — a premise that BDPD³ transforms into its central governance question: which signal (leading or lagging) should the regulator monitor? The lecture is a synthesis and programme statement; it does not provide a model of monitoring timing or signal selection.

Carlisle & Gruby (2019). Polycentric Systems of Governance: A Theoretical Model for the Commons. Policy Studies Journal, 47(4), 927–952. DOI: 10.1111/psj.12212

Polycentric Systems of Governance: A Theoretical Model for the Commons

AI Note: Carlisle and Gruby develop a theoretical model of polycentric governance systems for the commons, distinguishing attributes (definitional elements: multiple semiautonomous decision centres, overlapping jurisdiction, conflict-resolution capacity) from enabling conditions (additional institutional features for achieving functionality: trust, learning, redundancy). The model provides the most precise formalisation of polycentricity in the commons literature while acknowledging that systematic development of the concept, including its posited advantages, remains incomplete. BDPD’s engagement with this work spans multiple papers: paper_02 cites it to frame the nested-arena substrate against the polycentric governance literature; paper_03’s introduction uses it to characterise the polycentric extension of CPR theory and to motivate the lacuna the paper addresses — the temporal structure of the regulatory signal, which Carlisle and Gruby’s structural model does not address. The model is deliberately structural, specifying what a polycentric system is, not when or on what signal it acts — the gap that BDPD³ fills.

Heikkila et al. (2018). Bringing polycentric systems into focus for environmental governance. Environmental Policy and Governance, 28(4), 207–211. DOI: 10.1002/eet.1809

Bringing polycentric systems into focus for environmental governance

AI Note: Heikkila and colleagues introduce a special issue on polycentric environmental governance, defining polycentricity as multiple overlapping centres of decision-making interacting within an overarching set of rules. They identify the principal advantage of polycentric systems as the capacity to manage cross-scale environmental issues, while noting substantial diversity in the design and function of such systems globally. BDPD²’s introduction cites this definitional work to anchor its nested-World architecture against the polycentric governance literature. BDPD³’s introduction and discussion deploy the Heikkila framing to motivate the lacuna the paper addresses: the polycentric literature has thoroughly characterised which instruments to deploy but has said comparatively little about the temporal structure of the signal a regulator should monitor — specifically, whether it watches a leading or lagging indicator. The editorial is a programme statement and literature synthesis; it does not provide original empirical findings or model specific governance interventions.

Ahlström & Cornell (2017). Governance, polycentricity and the global nitrogen and phosphorus cycles. Environmental Science & Policy, 79, 54–65. DOI: 10.1016/j.envsci.2017.10.005

Governance, polycentricity and the global nitrogen and phosphorus cycles

AI Note: Ahlström and Cornell investigate the structural properties of governance for global nitrogen and phosphorus cycles using a mixed-methods approach that integrates polycentric theory with social network analysis of legal instruments. They find a loose, fragmented governance network in which EU legal instruments serve as key cross-scale gateways, and identify the Global Partnership on Nutrient Management as a nascent polycentric structure addressing governance gaps at the global level. BDPD³ cites this work as a real-world analogue for its cross-boundary pollution-cascade topology: the fragmented governance of N/P flows mirrors the A→B/C configuration, where one jurisdiction’s emissions cascade into downwind neighbours. BDPD³’s leading-indicator result — monitor capital rather than pollution — generates a falsifiable hypothesis for nutrient governance. The study maps institutional architecture, not agent-level dynamics; translating BDPD³’s signal-selection logic into policy recommendations for nutrient regimes would require additional institutional-process modelling.

Morrison et al. (2019). The black box of power in polycentric environmental governance. Global Environmental Change, 57, 101934. DOI: 10.1016/j.gloenvcha.2019.101934

The black box of power in polycentric environmental governance

AI Note: Morrison and colleagues examine power asymmetries in polycentric environmental governance, arguing that the polycentric literature has been slow to open the ‘black box of power.’ Even well-designed institutional instruments can be undermined by unequal power relations between decision centres — a corrective to the sometimes-optimistic tone of polycentric scholarship. BDPD²’s introduction cites this work alongside Baldwin et al. (2024) to justify its vignette format: the existing PG literature lacks temporal depth and power analysis, both of which the BDPD substrate can address in campaign-scale follow-ups. BDPD³’s introduction and discussion use the Morrison framing to motivate the leading/lagging signal distinction: even perfectly designed instruments fail when armed on the wrong signal, and power asymmetries compound the timing problem. The paper is a conceptual argument drawing on case-study synthesis; it does not provide a formal model of power dynamics.

Dorsch & Flachsland (2017). A Polycentric Approach to Global Climate Governance. Global Environmental Politics, 17(2), 45–64. DOI: 10.1162/glep_a_00400

A Polycentric Approach to Global Climate Governance

AI Note: Dorsch and Flachsland analyse global climate governance through an Ostromean polycentric lens, identifying four key mechanisms — self-organisation, site-specificity, experimentation, and trust-building — through which polycentric arrangements can enhance climate mitigation. They argue that the emerging global climate governance architecture is already polycentric in structure, even if not by design, and that deliberately strengthening polycentric features offers a viable pathway when centralised global agreements stall. BDPD² cites this work to licence the nested-substrate campaigns: a polycentric world architecture enables site-specific monitoring and experimentation across arenas. BDPD³’s discussion deploys the Dorsch-Flachsland framing to argue that site-specific monitoring must be assessed not only for coverage but for temporal position — is the monitored variable leading or lagging the governed process? The analysis is prescriptive and structural; it does not model agent behaviour or test governance interventions against a counterfactual, leaving the empirical gap that BDPD’s vignette-based and Seneca-engine experiments address.

Brunelli (2026). BDPD⁰ — A Computational Laboratory for Generative Agents, Commons Dilemmas and More. DOI: 10.5281/zenodo.20678100

BDPD⁰ — A Computational Laboratory for Generative Agents, Commons Dilemmas and More

Brunelli (2026). BDPD¹ — Governance: Cheap Talk and Sanctioning on a Fragile Commons. DOI: 10.5281/zenodo.20679157

BDPD¹ — Governance: Cheap Talk and Sanctioning on a Fragile Commons

Brunelli (2026). BDPD² — Governance of Nested Commons: Four Vignettes. DOI: 10.5281/zenodo.20679946

BDPD² — Governance of Nested Commons: Four Vignettes

Brunelli (2026). BDPD³ — Governing the Signal, Not the Symptom: Leading-Indicator Regulation Prevents the Seneca Cascade in a Polycentric Commons. DOI: 10.5281/zenodo.20680200

BDPD³ — Governing the Signal, Not the Symptom: Leading-Indicator Regulation Prevents the Seneca Cascade in a Polycentric Commons

Brunelli (2026). The Limits of Good Will — Ten Lectures on Strategic Interaction, Cooperation, and the Architecture of Decision. DOI: 10.5281/zenodo.20680327

The Limits of Good Will — Ten Lectures on Strategic Interaction, Cooperation, and the Architecture of Decision

Brunelli (2026). BDPD Research Notes — Six Dispatches from the Commons Laboratory. DOI: 10.5281/zenodo.20680823

BDPD Research Notes — Six Dispatches from the Commons Laboratory

Yin (2009). Case Study Research: Design and Methods. SAGE Publications. ISBN: 978-1412960991

Case Study Research: Design and Methods

AI Note: Yin’s methodological treatise on case-study research develops the logic of theoretical replication — using cases not for statistical generalisation but to test whether theoretical propositions hold across systematically varied conditions. BDPD¹ and BDPD² cite this to justify their N = 5 mini-pilot and vignette format: the load-bearing claims are qualitative (which cells preserve versus collapse) following theoretical replication logic rather than statistical generalisation. The Bayesian credible-interval supplements confirm that key contrasts are credible even at small N. The book addresses social-science methodology; it does not prescribe specific experimental designs for agent-based commons research.

Horton et al. (2026). Large Language Models as Simulated Economic Agents: What Can We Learn from Homo Silicus?. URL: https://arxiv.org/abs/2301.07543

Large Language Models as Simulated Economic Agents: What Can We Learn from Homo Silicus?

AI Note: Horton, Filippas, and Manning argue that LLMs are implicit computational models of humans — Homo silicus — that can be given endowments, information, and preferences, then studied via simulation. They replicate classic experiments (Charness-Rabin social preferences, Kahneman-Knetsch-Thaler fairness, Samuelson-Zeckhauser status-quo bias) and find qualitatively similar results to the originals, with divergences that are themselves generative of new research questions. BDPD⁰’s discussion and BDPD¹’s introduction both situate their LLM-agent experiments within the Homo silicus framework: LLMs are treated as empirical objects of behavioural inquiry — a complement, not a replacement, for heuristic agents. The framework’s empirical validation focuses on one-shot decisions and stated preferences; it has not been tested in the iterated, resource-coupled, strategic-extraction setting of the BDPD commons.

Brookins & DeBacker (2024). Playing games with GPT: What can we learn about a large language model from canonical strategic games?. Economics Bulletin, 44(1), 25–37. URL: https://ideas.repec.org/a/ebl/ecbull/eb-23-00457.html

Playing games with GPT: What can we learn about a large language model from canonical strategic games?

AI Note: Brookins and DeBacker have GPT-3.5 play dictator and prisoner’s dilemma games, comparing LLM decisions to human experimental baselines. The LLM displays strong fairness and cooperation tendencies: modal dictator allocation is a 50-50 split, and cooperation in the prisoner’s dilemma reaches ~65%, well above the human baseline of ~37%. These results suggest that LLM training distributions encode prosocial norms that dominate the self-interested Nash prediction. BDPD¹‘s D1 pilot and Lesson 7 of the mini-course cite this finding as independent evidence for the ’architecture effect’ — the transition from heuristic to LLM agents raises the cooperation index from 0.28 to 0.50 without any communication channel, a lift consistent with Brookins and DeBacker’s observation that LLMs exhibit intrinsic prosocial tendencies in canonical strategic games. The experiments use single-shot games with fixed payoff matrices, not the iterated, resource-coupled dilemmas with cumulative wealth dynamics that characterise the BDPD commons.

Aher et al. (2023). Using Large Language Models to Simulate Multiple Humans and Replicate Human Subject Studies. URL: https://proceedings.mlr.press/v202/aher23a.html

Using Large Language Models to Simulate Multiple Humans and Replicate Human Subject Studies

AI Note: Gati and colleagues introduce the Turing Experiment (TE) methodology for evaluating how well language models simulate human behaviour across representative subject samples, replicating classic economic, psycholinguistic, and social psychology experiments. They find that LLM-based replications, while qualitatively faithful, exhibit systematic distortions including a hyper-accuracy effect — LLMs produce less noisy, more deterministic responses than human subjects. BDPD¹ and BDPD² cite this methodology as a caveat for their single-model LLM results: the preservation rate observed with DeepSeek-flash should be read as an upper bound rather than a point estimate, and the architectural cooperation lift may reflect this family’s training distribution. The TE framework focuses on replicating well-established between-subject effects; it does not address within-subject strategic dynamics across iterated interactions in a resource-coupled commons.

Balliet (2010). Communication and Cooperation in Social Dilemmas: A Meta-Analytic Review. Journal of Conflict Resolution, 54(1), 39–57. DOI: 10.1177/0022002709352443

Communication and Cooperation in Social Dilemmas: A Meta-Analytic Review

AI Note: Balliet’s meta-analysis of 45 effect sizes from public-goods and prisoner’s-dilemma experiments reports a large overall positive effect of communication on cooperation (Cohen’s d = 1.01), with face-to-face discussion (d = 1.21) substantially stronger than written messages (d = 0.46). The communication–cooperation relationship is stronger in larger groups — counter to the Olson intuition — and does not depend on whether communication occurs before or during iterated dilemmas. BDPD¹ uses this meta-analytic baseline as the foil for its cheap-talk null: on the fragile BDPD cliff, written-message communication produces only d = 0.17, a negligible effect against the d = 0.46 benchmark for the same modality, isolating architecture (LLM vs heuristic) as the dominant driver. The meta-analysis pools exclusively human-subject studies; none of the included experiments involved LLM agents or commons with asymmetric collapse dynamics.

Agranov (2025). Communication in Games. North Holland. ISBN: 9780443317583

Communication in Games

AI Note: Agranov surveys the experimental literature on pre-play cheap-talk communication across game types, organising findings by structural prerequisites: communication reliably shifts outcomes toward Pareto-efficient equilibria in coordination games with multiple rankable equilibria, but its effectiveness is gated by preference alignment, group size, and message richness. In mixed-motive games lacking rankable equilibria, communication has limited success. BDPD¹’s cheap-talk null on a fragile commons (Cohen’s d = 0.17) operationalises exactly the structural precondition Agranov identifies: the BDPD cliff configuration has no alternative equilibrium for talk to select, so the survey predicts a null — and the pilot empirically delivers one. To our reading, the survey covers only human-subject experiments in laboratory settings; how its structural prerequisites translate to LLM–agent populations with different cognitive architectures remains an open empirical question.

van Klingeren & Buskens (2024). Graduated sanctioning, endogenous institutions and sustainable cooperation in common-pool resources: An experimental test. Rationality and Society, 36(2), 183–229. DOI: 10.1177/10434631231219608

Graduated sanctioning, endogenous institutions and sustainable cooperation in common-pool resources: An experimental test

AI Note: Van Klingeren and Buskens compare graduated versus strict mutual sanctioning in a common-pool resource laboratory experiment, also distinguishing exogenously imposed from endogenously chosen sanctioning regimes. They find that graduated sanctioning outperforms strict sanctioning in the long run by reducing retaliatory responses and increasing perceived legitimacy — the first low penalty does not provoke counter-punishment as an immediate strict penalty does. BDPD¹’s D3 ladder-sweep pilot operationalises this finding: three graduated schedules preserve the commons in ≥4/5 LLM-driven seeds, while flat sanctions preserve far fewer. The human-subject experiment confirms the mechanism BDPD¹ then tests on an LLM-agent substrate. The study uses a standard CPR game with human subjects; it does not test graduated sanctioning with LLM agents or on a substrate with irreversible Seneca-style collapse.

Couto et al. (2020). Governance of risky public goods under graduated punishment. Journal of Theoretical Biology, 505, 110423. DOI: 10.1016/j.jtbi.2020.110423

Governance of risky public goods under graduated punishment

AI Note: Couto, Pacheco, and Santos model the governance of risky public goods under graduated punishment using evolutionary game theory. They show that escalation schedules promote cooperation at lower average severity than flat sanctions — a graduated ladder peaking at a moderate level can outperform a higher flat sanction because the escalation shape, not the terminal value, drives the evolutionary dynamics. The mechanism: graduated schedules create a behavioural gradient that deters continued violation without provoking the retaliatory responses that flat severe sanctions trigger. BDPD¹’s D3 ladder-sweep pilot operationalises this finding: three graduated schedules — [1,2,4], [1,3,10], [1,5,25] — preserve the commons in ≥4/5 seeds, while a flat-5 control preserves only 1/5, confirming that escalation shape carries the effect. The model is an analytical evolutionary framework with homogeneous agents and fixed payoff structures; it does not incorporate the strategic heterogeneity or institutional complexity of the BDPD nested-world substrate.

Li et al. (2026). Mathematical Modeling of Common-Pool Resources: A Comprehensive Review of Bioeconomics, Strategic Interaction, and Complex Adaptive Systems. URL: https://arxiv.org/abs/2602.03129

Mathematical Modeling of Common-Pool Resources: A Comprehensive Review of Bioeconomics, Strategic Interaction, and Complex Adaptive Systems

AI Note: Li and colleagues provide a comprehensive review of mathematical modeling of common-pool resources, tracing the intellectual trajectory from deterministic bioeconomic models through game-theoretic formalisations — including the Ostrom Turn toward institutional realism — to evolutionary game theory, stochastic differential equations, and agent-based computational economics. BDPD¹’s introduction cites this review to position heuristic agents as the dominant paradigm in the CPR agent-based modelling literature, against which LLM agents are offered as a complement, not a replacement. The review synthesises the formal modelling tradition that BDPD extends by adding the agent-architecture dimension. As a review, it organises existing literature rather than producing new model results or governance findings.

Mavi & Quérou (2021). Common Pool Resource Management and Risk Perceptions. URL: https://ideas.repec.org/p/fae/wpaper/2021.02.html

Common Pool Resource Management and Risk Perceptions

AI Note: Mavi and Quérou introduce a non-cooperative game model of common-pool resource management under heterogeneous risk-perception biases, showing that the type of bias (overestimation versus underestimation) and resource quality before and after a regime shift have first-order importance on conservation patterns. Under non-uniform biases, the intra-group structure qualitatively affects conservation, and unbiased agents may react non-monotonically to polarisation. BDPD²’s V4 vignette cites this alongside Centola (2018) to support the analytical prediction that heterogeneous risk perception produces non-monotonic behavioural adjustments — consistent with the active-contagion hypothesis for the V4 purity drop. The model is an analytical game-theoretic framework with fixed payoff structures; it does not incorporate the strategic extraction dynamics or institutional meta-agents of the BDPD nested substrate.

Ritchie et al. (2023). Rate-Induced Tipping in Natural and Human Systems. Earth System Dynamics, 14(3), 669–683. DOI: 10.5194/esd-14-669-2023

Rate-Induced Tipping in Natural and Human Systems

AI Note: Ritchie and colleagues examine rate-induced tipping in the context of climate and ecological systems, characterising the conditions under which the speed of change — rather than the absolute magnitude — drives a system across a critical threshold. BDPD²’s conclusions cite this alongside the rate-induced boundary identified in BDPD³, where the governance sampling interval dt determines whether the regulator keeps pace with capital dynamics. The rate-induced framing provides the conceptual bridge between BDPD³’s Seneca-engine results and broader environmental governance: the binding constraint is not only which signal to monitor but how fast to sample it. The paper addresses ecological systems without strategic agents or governance institutions.

Gao et al. (2023). S3: Social-Network Simulation System with Large Language Model-Empowered Agents. SSRN Electronic Journal. DOI: 10.2139/ssrn.4607026

S3: Social-Network Simulation System with Large Language Model-Empowered Agents

AI Note: Gao and colleagues present S3, a social-network simulation system using LLM-empowered agents that emulate human-like behaviour in emotion, attitude, and interaction. By fine-tuning and prompt-engineering LLM agents to perceive their informational environment, the system reproduces population-level phenomena — information propagation, attitude shifts, emotion contagion — with promising accuracy against real-world social network data. BDPD²’s introduction situates its work within the LLM-driven social-simulation taxonomy, citing S3 as representative of Society Simulation where the World layer carries meta-agents — the register toward which BDPD’s nested-World architecture reaches. S3 simulates social network dynamics on a fixed graph topology; it does not incorporate resource dynamics, strategic extraction, or institutional governance mechanisms, limiting direct translation to commons dilemmas.

Bito et al. (2026). Large Language Models Exhibit Normative Conformity. URL: https://arxiv.org/abs/2604.19301

Large Language Models Exhibit Normative Conformity

AI Note: Bito and colleagues test six LLMs on tasks designed to distinguish informational conformity (changing one’s judgment because others’ answers provide useful information) from normative conformity (changing one’s judgment to avoid conflict or gain social acceptance). They find that up to five of six LLMs exhibit normative conformity, and that manipulating social context — peer endorsements, shared attributes — can redirect which speaker an LLM conforms to, raising the risk of manipulation in multi-agent LLM systems. BDPD²‘s Vignette 4 cites this work as a candidate mechanism for the ’purity drop’ observed under coercive exclusion: LLM conformists may co-breach a pact alongside defectors due to normative conformity triggered by observing the defector’s visible extraction. The paper tests LLMs in dyadic or small-group judgment tasks, not in the open-ended strategic commons setting of BDPD, so whether normative conformity scales to resource dilemmas with cumulative payoffs remains unverified.

Bojić et al. (2025). An Agent-Based Simulation of Politicized Topics Using Large Language Models: Algorithmic Personalization and Polarization on Social Media. Chinese Political Science Review. DOI: 10.1007/s41111-025-00326-x

An Agent-Based Simulation of Politicized Topics Using Large Language Models: Algorithmic Personalization and Polarization on Social Media

AI Note: Bojić and colleagues introduce RecSysLLMsP, a simulation framework with 100 LLM-based agents grounded in psychometric and demographic data to examine how algorithmic personalisation interacts with language generation to influence polarisation on social media. Moderate personalisation maximises engagement; full personalisation reduces diversity and amplifies both structural and affective polarisation (modularity Q: 0.22→0.68). BDPD²’s V4 vignette cites this work alongside Bito et al. as candidate evidence for the normative-conformity mechanism: LLM agents adjust toward locally dominant behaviour even when informational signals would not warrant the change, consistent with the purity drop observed under coercive exclusion. The study models polarisation in a social-media context with recommender-driven content feeds, not in an extraction dilemma with resource feedback; whether the same conformity dynamics survive in a CPR setting with cumulative wealth consequences remains untested.

Mou et al. (2026). From Individual to Society: A Survey on Social Simulation Driven by Large Language Model-Based Agents. ACM Computing Surveys, 58(11). DOI: 10.1145/3800683

From Individual to Society: A Survey on Social Simulation Driven by Large Language Model-Based Agents

AI Note: Mou and colleagues propose a taxonomy of LLM-driven social simulation spanning Individual, Interaction, Scenario, and Society registers, distinguishing simulations by the number of agents, interaction structure, institutional context, and presence of meta-agents. BDPD⁰’s discussion and BDPD²’s introduction use this taxonomy to situate the BDPD platform: heuristic agents on the Arena sit in a lower register, while LLM agents with pacts, treaties, and World-level meta-agents approach the Society Simulation register. The taxonomy is descriptive and classificatory; it does not prescribe implementation details or evaluate which simulation register is appropriate for which research question.

Mahmoud et al. (2014). A Review of Norms and Normative Multiagent Systems. The Scientific World Journal, 2014, 684587. DOI: 10.1155/2014/684587

A Review of Norms and Normative Multiagent Systems

AI Note: Mahmoud and colleagues review the state of research on norms and normative multiagent systems, proposing a norm life-cycle model covering norm emergence, recognition, adoption, and enforcement. The review synthesises definitions, classifications, and architectures for embedding norms in agent societies. BDPD²’s platform section cites this work to frame its treaties as norms in the normative-MAS sense: treaties are observed by meta-agents and enforced through sanctions but impose no hard constraint on player choice — distinguishing them from hard-coded rules. The normative-MAS framing provides the conceptual vocabulary for BDPD’s graduated-sanction architecture. The review addresses software-agent norms with rule-based compliance mechanisms; it does not consider LLM agents whose norm sensitivity is an emergent property of training rather than an explicit compliance module.

Thiel (2023). Polycentric Governing and Polycentric Governance. Oxford University Press. DOI: 10.1093/oso/9780192866837.003.0005

Polycentric Governing and Polycentric Governance

AI Note: Thiel and colleagues provide a systematic analysis of polycentric governance, clarifying conceptual elements and distinguishing positive analysis from normative advocacy. They argue that PG scholarship has often assumed rather than empirically tested the desirability of polycentric arrangements. BDPD²’s platform section cites this alongside Ostrom (1990, 2010) to define its Bloomington-school polycentric frame, while sharing Thiel’s concern that polycentric governance requires rigorous empirical testing rather than assumption. The BDPD vignette format — comparing institutional mechanisms under both heuristic and LLM agents — operationalises exactly the kind of empirical PG testing Thiel advocates. The paper is conceptual and meta-theoretical; it does not present original empirical findings or a simulation framework.

Baldwin et al. (2024). Empirical Research on Polycentric Governance: Critical Gaps and a Framework for Studying Long-Term Change. Policy Studies Journal, 52(2), 319–348. DOI: 10.1111/psj.12518

Empirical Research on Polycentric Governance: Critical Gaps and a Framework for Studying Long-Term Change

AI Note: Baldwin and colleagues conduct a systematic review of the empirical polycentric governance (PG) literature, finding that most studies provide short-term snapshots that fail to address change and evolution over time. The review identifies both positive and negative features associated with PG but reveals limited cumulation of knowledge about why PG works in some cases and not in others. They propose a COOF (Context–Operations–Outcomes–Feedbacks) framework to guide longitudinal PG research. BDPD² cites this review to justify its vignette-based N = 5 survey format: the prevailing empirical PG literature lacks temporal depth, and the BDPD substrate enables exactly the campaign-scale longitudinal follow-ups that Baldwin et al. identify as missing. The COOF framework offers a conceptual scaffold for designing such follow-ups within the BDPD nested-world architecture.

Lippe et al. (2019). Using Agent-Based Modelling to Simulate Social-Ecological Systems Across Scales. GeoInformatica, 23(2), 269–298. DOI: 10.1007/s10707-018-00337-8

Using Agent-Based Modelling to Simulate Social-Ecological Systems Across Scales

AI Note: Lippe and colleagues survey the state of agent-based modelling for social-ecological systems across scales, arguing that current ABM architecture generally does not translate cleanly from village or landscape to regional or global contexts. They identify a structural gap: models are usually scoped by pragmatic considerations that cut across dynamical boundaries, and the relevant scale of analysis is often not known a priori. BDPD² cites this to position its nested-World substrate as addressing exactly this gap — the BDPD World object wraps multiple arenas, resource links, and meta-agents in a way that lifts ABM from single-arena to multi-scale governance. The paper is a conceptual review and thought-piece; it does not present a working multi-scale ABM implementation.

Fischbacher et al. (2001). Are people conditionally cooperative? Evidence from a public goods experiment. Economics Letters, 71(3), 397–404. DOI: 10.1016/S0165-1765(01)00394-9

Are people conditionally cooperative? Evidence from a public goods experiment

AI Note: Fischbacher, Gächter, and Fehr use the strategy method to elicit individual contribution schedules in a one-shot public goods game, classifying subjects into behavioural types. They find that approximately 50% are conditional cooperators (matching their contribution to others’), about 30% are free-riders, and the remainder are distributed across hump-shaped and other patterns. BDPD²’s cross-cutting analysis cites this as the canonical human benchmark for the heterogeneity that matters: conditional cooperators adjust to peer behaviour, while free-riders do not — analogous to the heuristic-vs-LLM architecture divide where rule-bound agents produce flat baselines while signal-adaptive agents decode institutional context. The classification uses a one-shot strategy-method design; the types, while replicable, are only moderately stable across measurement waves, as Volk et al. (2012) later showed.

Volk et al. (2012). Temporal stability and psychological foundations of cooperation preferences. Journal of Economic Behavior & Organization, 81(2), 664–676. DOI: 10.1016/j.jebo.2011.10.006

Temporal stability and psychological foundations of cooperation preferences

AI Note: Volk, Thöni, and Ruigrok conduct a longitudinal study of cooperative types — conditional cooperators, free-riders, and others — across three measurement waves spanning five months, finding that types are only moderately stable: about half of subjects retain the same classification, and test-retest agreement is fair (κ ≈ 0.3–0.5). BDPD²’s cross-cutting analysis and the mini-course Lesson 1 cite this to argue that architecture-level variation (LLM vs heuristic) operates on a categorically different dimension than within-human type drift. BDPD’s finding that governance effects invert depending on agent architecture is thus not merely an extreme case of within-population heterogeneity — it is a fundamentally different source of variation. The study addresses human subjects in standard public goods games; it does not consider LLM agents.

Janssen et al. (2010). Lab Experiments for the Study of Social-Ecological Systems. Science, 328(5978), 613–617. DOI: 10.1126/science.1183532

Lab Experiments for the Study of Social-Ecological Systems

AI Note: Janssen, Holahan, Lee, and Ostrom use a spatial common-pool resource experiment to study how communication and punishment — alone, combined, and in different temporal orders — interact in commons management. Their headline finding is that communication alone sustains cooperation about as effectively as costly punishment, and that introducing costly punishment after a successful communication round actively erodes the cooperative agreements that communication had produced. BDPD²’s cross-cutting analysis cites this as the closest methodological precedent for its governance vignettes, which vary the decision-making substrate (heuristic vs LLM) rather than just the institutional levers. BDPD’s mini-course Lesson 1 uses the finding to argue that institutional choice in CPR settings is sensitive to who plays and to the order of institutional moves. The experiment uses human subjects with a spatial renewable resource; the BDPD extension to LLM agents on a logistic or Seneca engine with irreversible collapse represents a novel experimental dimension.

Burlando & Guala (2005). Heterogeneous agents in public goods experiments. Experimental Economics, 8(1), 35–54. DOI: 10.1007/s10683-005-0436-4

Heterogeneous agents in public goods experiments

AI Note: Burlando and Guala explore agent heterogeneity in public goods experiments by classifying subjects into free-riders, cooperators, and reciprocators, then re-playing the game in homogeneous groups. They find that free-rider groups rapidly decay to zero contribution while cooperative and reciprocating groups sustain high, stable contributions — confirming that type composition, not just group size or payoff structure, is an active causal variable driving the decay phenomenon. BDPD²’s cross-cutting analysis cites this as a within-human parallel to the heuristic-vs-LLM architecture divide: rule-bound agents produce flat baselines (like sorted free-rider groups that decay predictably), while signal-adaptive agents respond to institutional context in both directions. The BDPD mini-course Lesson 1 uses this study alongside Fischbacher et al. (2001) to establish that mixed populations behave in ways not linearly interpolable from pure types. The experiment addresses preference-type heterogeneity within a single species, not the radical heterogeneity of agent architecture that BDPD studies.

Sally (1995). Conversation and cooperation in social dilemmas: A meta-analysis of experiments from 1958 to 1992. Rationality and Society, 7(1), 58–92. DOI: 10.1177/1043463195007001004

Conversation and cooperation in social dilemmas: A meta-analysis of experiments from 1958 to 1992

AI Note: Sally’s meta-analysis of 130 experimental treatments published between 1958 and 1992 — the first quantitative synthesis of the communication-and-cooperation literature — finds that the mere presence of discussion raises cooperation rates by more than 45 percentage points, an enormous effect. BDPD’s mini-course Lesson 1 uses this as the classical benchmark against which the BDPD¹ cheap-talk null is measured: the +45pp effect among humans, across decades of experiments, makes the d = 0.17 result among LLM agents a striking anomaly. Lesson 2 uses Sally’s finding that group size barely registers in the full sample — a log-linear effect of ~7% defection per doubling — to argue against Olson’s qualitative discontinuity claim. The meta-analysis pools studies with heterogeneous designs and exclusively human subjects.

Camerer (2003). Behavioral Game Theory: Experiments in Strategic Interaction. Princeton University Press. ISBN: 978-0-691-09039-9

Behavioral Game Theory: Experiments in Strategic Interaction

AI Note: Camerer’s Behavioral Game Theory is the canonical textbook synthesising experimental results on strategic interaction, covering dictator and ultimatum games, coordination, bargaining, learning, and cheap talk across hundreds of laboratory studies. It establishes the empirical regularities — overcontribution in public goods, the power of face-to-face communication, bounded rationality in strategic reasoning — that became the baseline against which agent-based and LLM-driven experiments are now compared. BDPD’s mini-course uses Camerer as the standard reference across multiple lessons: Lesson 1 for the cheap-talk record, Lesson 2 for group-size effects and ‘cloning heuristics’ (the finding that subjects extrapolate from small-group experience rather than compute marginal incentives), and Lesson 4 for counter-punishment escalation dynamics. The book synthesises human-subject experiments exclusively; its empirical regularities describe Homo sapiens, not Homo silicus, making it the natural foil for the architecture-effect decomposition that structures BDPD’s experimental programme.

Yamagishi (1986). The provision of a sanctioning system as a public good. Journal of Personality and Social Psychology, 51(1), 110–116. DOI: 10.1037/0022-3514.51.1.110

The provision of a sanctioning system as a public good

AI Note: Yamagishi demonstrates experimentally that the provision of a sanctioning system is itself a second-order public good: individuals free-ride on the contribution to the sanctioning infrastructure, and groups that voluntarily adopt sanctioning systems sustain higher cooperation than those that do not. BDPD⁰’s discussion cites this finding to note that sanctioning is not free — the BDPD platform deliberately isolates structural dynamics before adding the second-order governance layer in BDPD¹. The graduated-sanction pilots in BDPD¹ test whether institutional (exogenous) sanctioning can short-circuit the second-order problem that Yamagishi’s endogenous-sanctioning design identifies. The experiment uses human subjects with a standard public goods game; it does not address LLM agents or irreversible collapse.

Ostrom et al. (1992). Covenants With and Without a Sword: Self-Governance Is Possible. The American Political Science Review, 86(2), 404–417. DOI: 10.2307/1964229

Covenants With and Without a Sword: Self-Governance Is Possible

AI Note: Ostrom, Walker, and Gardner test the Hobbesian claim that ‘covenants, without the sword, are but words’ through a common-pool resource experiment with three treatments: covenants alone (one-shot and repeated communication), swords alone (repeated sanctioning opportunities), and covenants combined with an internal sword (communication followed by sanctioning). They demonstrate that self-governance is possible — communication alone substantially improves CPR outcomes, and combining communication with sanctioning produces the strongest results, directly challenging the Hobbesian presumption that external enforcement is necessary. BDPD⁰’s discussion cites this as the canonical demonstration that enforcement, not communication alone, is the load-bearing governance component on a commons — the premise BDPD¹’s graduated-sanction pilots operationalise. BDPD¹ explicitly frames its cheap-talk-plus-sanctioning architecture as the second column of Ostrom’s design-principles table. The experiment uses human subjects with a renewable CPR; it does not model the irreversible Seneca cliff or LLM agents.

Isaac & Walker (1988). Communication and Free-Riding Behavior: The Voluntary Contribution Mechanism. Economic Inquiry, 26(4), 585–608. DOI: 10.1111/j.1465-7295.1988.tb01519.x

Communication and Free-Riding Behavior: The Voluntary Contribution Mechanism

AI Note: Isaac and Walker examine the role of communication in the voluntary contribution mechanism (VCM), the canonical public-goods experimental paradigm. They find that without communication, contributions decay toward zero with repetition, but that communication significantly improves group optimality, with robustness tested in increasingly complex environments. BDPD’s mini-course Lesson 1 uses this as the foundational demonstration that communication works in the VCM paradigm — the baseline against which the BDPD¹ null (LLM chat, d = 0.17) is measured. The VCM uses four-player groups, ten-period interactions, and fixed marginal per-capita returns — a substantially different setting from the BDPD cliff configuration with six agents, irreversible collapse, and logistic resource regeneration. The paper established the experimental paradigm; it does not address LLM agents or resource-coupled dynamics.

Oprea et al. (2014). Continuous time and communication in a public-goods experiment. Journal of Economic Behavior & Organization, 108, 212–223. DOI: 10.1016/j.jebo.2014.09.012

Continuous time and communication in a public-goods experiment

AI Note: Oprea, Charness, and Friedman run a public-goods experiment comparing discrete-time versus continuous-time interaction, each crossed with pre-play free-form text chat. In continuous time with communication, contributions approach 100% of the endowment — the highest cooperation rates ever recorded in a VCM experiment. Their motivating observation was that most real-world public goods have a real-time aspect, yet the entire experimental literature had operationalised provision as discrete decision points. BDPD’s mini-course Lesson 1 cites this as the contemporary state of the art on cheap talk in public-goods games — the strongest possible human baseline against which the BDPD¹ null (LLM chat, d = 0.17) is measured. The experiment uses human subjects with a standard linear VCM; it does not model resource dynamics or the irreversible collapse that characterises the BDPD cliff.

Loomis (1959). Communication, the Development of Trust, and Cooperative Behavior. Human Relations, 12(4), 305–315. DOI: 10.1177/001872675901200402

Communication, the Development of Trust, and Cooperative Behavior

AI Note: Loomis’ foundational experiment demonstrates that communication builds trust and sustains cooperative behaviour in a social dilemma, conducted at NYU under Morton Deutsch’s direction before the Voluntary Contribution Mechanism existed as an experimental paradigm. His framework identifies four prerequisites for trust-based cooperation — commitment, mutual dependence, perceived mutual dependence, and mutually perceived interdependence — and argues that communication directly engineers the latter two. BDPD’s mini-course Lesson 1 opens with Loomis as the historical anchor of sixty years of evidence that communication robustly enhances cooperation among humans. The BDPD¹ null (LLM chat, d = 0.17) is then framed as a striking departure from this sixty-year consensus. The experiment was a small-scale trust game, not a repeated commons dilemma with resource dynamics and collapse thresholds.

Bochet et al. (2006). Communication and punishment in voluntary contribution experiments. Journal of Economic Behavior & Organization, 60(1), 11–26. DOI: 10.1016/j.jebo.2003.06.006

Communication and punishment in voluntary contribution experiments

AI Note: Bochet, Page, and Putterman compare three communication forms — face-to-face, anonymous text chat, and structured numerical announcements — crossed with costly punishment in a public goods experiment. They find that anonymous text chat is nearly as effective as face-to-face communication, while numerical cheap talk has no effect. Punishment increases contributions but reduces efficiency due to its cost and because of ‘perverse punishment’ directed at high contributors. BDPD’s mini-course Lesson 1 uses this study to map the communication-media gradient: once vocal and visual cues are stripped away, text-based communication still works — a finding that makes the BDPD¹ null (LLM written chat, d = 0.17) all the more striking, since Bochet’s human subjects achieved near-face-to-face cooperation through the same written modality. The experiment uses human subjects with standard VCM payoffs; the generalisation to LLM agents on a fragile, asymmetric commons remains open.

Brosig et al. (2003). The Effect of Communication Media on Cooperation. German Economic Review, 4(2), 217–241. DOI: 10.1111/1468-0475.00080

The Effect of Communication Media on Cooperation

AI Note: Brosig, Weimann, and Ockenfels compare the effect of communication media on cooperation in public goods experiments, gradually peeling back components of face-to-face communication — physical co-presence, visual cues, aural cues, bidirectionality — to identify which channels drive the cooperation-enhancing effect. They find that successful cooperation is attributable to the opportunity to coordinate behaviour, and that both the level and stability of cooperation significantly interact with the communication medium, even though the content of communication is remarkably similar across treatments. BDPD’s mini-course Lesson 1 cites this study as a key piece of the communication-media gradient that contextualises the BDPD¹ null: since even stripped-down channels can sustain cooperation among humans, the failure of the same channel among LLMs isolates the agent architecture as the operative variable. The study uses human subjects; its channel-decomposition logic has not been replicated with LLM agents.

McGinnis & Ostrom (2014). Social-ecological system framework: initial changes and continuing challenges. Ecology and Society, 19(2), 30. DOI: 10.5751/ES-06387-190230

Social-ecological system framework: initial changes and continuing challenges

AI Note: McGinnis and Ostrom present the revised social-ecological system (SES) framework, summarising changes to the original Ostrom (2007) formulation and addressing remaining ambiguities. The framework provides a common vocabulary for analysing governance systems, resource units, actors, and their interactions across diverse empirical settings. BDPD² cites this alongside Lippe et al. (2019) to establish the structural gap its nested-World substrate addresses: most SES-framework applications operate at a single scale, while BDPD’s World object enables multi-arena, cross-scale governance analysis. The SES framework is a diagnostic tool for empirical case studies, not an operational simulation framework; the BDPD platform translates its conceptual categories into executable agent-based experiments.

Cash et al. (2006). Scale and cross-scale dynamics: governance and information in a multilevel world. Ecology and Society, 11(2), 8. DOI: 10.5751/ES-01759-110208

Scale and cross-scale dynamics: governance and information in a multilevel world

AI Note: Cash and colleagues synthesise evidence on scale and cross-scale dynamics in environmental governance, arguing that the interplay between institutions at multiple levels determines the success or failure of managing cross-boundary environmental problems. They propose that co-management structures and conscious boundary management — including knowledge co-production, mediation, translation, and negotiation across scale-related boundaries — may facilitate solutions to complex problems that have historically resisted resolution. BDPD² cites this work alongside Ahlström and Cornell (2017) as part of the real-world evidence for cross-boundary governance challenges, motivating its Road-A proposal to route hidden pollution state into a world-level enforcer on the Seneca substrate. The paper is a synthesis of case studies and conceptual frameworks; it does not provide a formal model of cross-scale dynamics or test specific governance interventions against a counterfactual.

Bikhchandani et al. (1992). A Theory of Fads, Fashion, Custom, and Cultural Change as Informational Cascades. Journal of Political Economy, 100(5), 992–1026. URL: http://www.jstor.org/stable/2138632

A Theory of Fads, Fashion, Custom, and Cultural Change as Informational Cascades

An informational cascade occurs when it is optimal for an individual, having observed the actions of those ahead of him, to follow the behavior of the preceding individual without regard to his own information. We argue that localized conformity of behavior and the fragility of mass behaviors can be explained by informational cascades.

AI Note: Bikhchandani, Hirshleifer, and Welch (BHW) develop a model of informational cascades in which rational agents, observing the choices but not the private signals of predecessors, may converge on an incorrect action — and once a cascade forms, it becomes self-reinforcing and fragile only to the arrival of new public information. The model explains fads, fashion, and sudden shifts in social behaviour as products of Bayesian rationality, not irrationality. BDPD’s mini-course Lesson 8 — “More Information Equals Better Decisions” — uses BHW as the canonical theoretical statement that public observability of others’ choices can produce systematically wrong collective outcomes, directly motivating the platform’s finding in paper_00 that observation noise improves welfare by breaking the informational channel that aggressive agents exploit. The BHW model assumes a fixed signal accuracy; it does not model the endogenous deterioration of signal quality under learning or strategic feedback.

Banerjee (1992). A Simple Model of Herd Behavior. The Quarterly Journal of Economics, 107(3), 797-817. DOI: 10.2307/2118364

A Simple Model of Herd Behavior

We analyze a sequential decision model in which each decision maker looks at the decisions made by previous decision makers in taking her own decision. This is rational for her because these other decision makers may have some information that is important for her. We then show that the decision rules that are chosen by optimizing individuals will be characterized by herd behavior; i.e., people will be doing what others are doing rather than using their information. We then show that the resulting equilibrium is inefficient.

AI Note: Banerjee develops a sequential decision model in which each agent observes the choices of predecessors and decides whether to follow their own private information or join the emerging pattern. The model shows that rational agents may ignore their private signals even when the aggregate information strongly favours another choice, producing inefficient ‘herd behaviour.’ Banerjee identifies a ‘herd externality’: each agent’s decision to follow the herd deprives later agents of the informational content of her private signal. BDPD’s mini-course Lesson 8 cites Banerjee alongside Bikhchandani-Hirshleifer-Welch (1992) to anchor the informational-cascade mechanism, emphasising reputational reinforcement: agents who care about how their choices are evaluated have an additional incentive to follow the herd, beyond the purely informational logic. The model assumes a fixed signal structure and sequential choice; it does not incorporate endogenous signal quality or resource feedback, dimensions central to the BDPD commons substrate.

Anderson & Holt (1997). Information Cascades in the Laboratory. The American Economic Review, 87(5), 847–862. URL: http://www.jstor.org/stable/2951328

Information Cascades in the Laboratory

When a series of individuals with private information announce public predictions, initial conformity can create an “information cascade” in which later predictions match the early announcements. This paper reports an experiment in which private signals are draws from an unobserved urn. Subjects make predictions in sequence and are paid if they correctly guess which of two urns was used for the draws. If initial decisions coincide, then it is rational for subsequent decision makers to follow the established pattern, regardless of their private information. Rational cascades formed in most periods in which such an imbalance occurred.

AI Note: Anderson and Holt test the rational information-cascade model of Bikhchandani, Hirshleifer, and Welch (1992) in a laboratory experiment. Subjects draw private signals from an unobserved urn and make sequential public predictions about which of two urns was selected. When early decisions coincide, it becomes rational for later subjects to ignore their private signals and follow the established pattern, even when that pattern is incorrect. The experiment confirms that rational cascades form in most periods where signal imbalance occurs. BDPD’s mini-course Lesson 8 — “More Information Equals Better Decisions” — cites this as the canonical experimental demonstration that public observability of others’ choices can produce systematic errors, directly motivating the BDPD platform’s finding that observation noise on the commons stock improves welfare by breaking informational cascades that would otherwise concentrate aggressive extraction. The experiment uses a fixed signal structure (2/3 accuracy) rather than the fluid, endogenous information environment of an agent-based commons.

Çelen & Kariv (2004). Distinguishing Informational Cascades from Herd Behavior in the Laboratory. American Economic Review, 94(3), 484–498. DOI: 10.1257/0002828041464461

Distinguishing Informational Cascades from Herd Behavior in the Laboratory

AI Note: Çelen and Kariv conduct a laboratory experiment that distinguishes informational cascades from herd behaviour by eliciting subjects’ beliefs. In an informational cascade, agents rationally infer that the crowd possesses superior information and ignore their private signals; in herd behaviour, agents follow the crowd because deviation is costly regardless of their private beliefs. By adding a setup with continuous signals and discrete actions to the Anderson-Holt paradigm, they show that the two mechanisms are empirically separable and have different policy implications. BDPD’s mini-course Lesson 8 cites this distinction as central to the cascade-governance question: if conformity in multi-agent systems is informational, the remedy is revealing private information; if it is reputational, the remedy is changing incentives. The BDPD platform’s noise-finding — that observation opacity improves welfare — speaks to the informational channel. The experiment uses human subjects with a fixed urn-and-ball signal structure, not the endogenous information environment of an agent-based commons.

Friedman (1971). A Non-cooperative Equilibrium for Supergames. The Review of Economic Studies, 38(1), 1-12. DOI: 10.2307/2296617

A Non-cooperative Equilibrium for Supergames

AI Note: Friedman proves the first formal folk theorem for infinitely repeated games (supergames): any outcome that gives each player more than their minimax payoff can be sustained as a subgame-perfect equilibrium using trigger strategies, provided the discount factor is sufficiently high. The proof is constructive — cooperate while all have cooperated; revert to minimax punishment forever upon any deviation. BDPD’s mini-course Lesson 9 — “If We Meet Again, We Cooperate” — uses Friedman as the historical starting point of the folk theorem tradition, then traces the expansion that revealed both the power and the emptiness of the result: that repetition opens a vast space of equilibria including full cooperation and full mutual exploitation. The theorem assumes infinite repetition with perfect monitoring; the BDPD cliff configuration operates under finite horizons with irreversible collapse, where the trigger-strategy logic does not directly apply.

Aumann & Shapley (2013). Long-Term Competition — A Game-Theoretic Analysis. Annals of Economics and Finance, 14(2(B)), 609–622. URL: https://api.semanticscholar.org/CorpusID:12445032

Long-Term Competition — A Game-Theoretic Analysis

AI Note: Aumann and Shapley’s foundational working paper extends the folk theorem of repeated games to a broader class of informational settings, showing that any feasible and individually rational payoff vector can be supported as a Nash equilibrium of the infinitely repeated game under the limiting-average payoff criterion. The paper demonstrates that cooperative outcomes can emerge from purely noncooperative play sustained by the threat of punishment — but also that this multiplicity eliminates predictive power. BDPD’s mini-course Lesson 9 — “If We Meet Again, We Cooperate” — uses this work to anchor the folk theorem’s double-edged nature: almost any outcome is an equilibrium, including full mutual exploitation. The paper’s focus on the limiting-average case with perfect monitoring means it abstracts away from the strategic uncertainty and finite horizons that characterise the BDPD cliff configuration.

Kreps et al. (1982). Rational cooperation in the finitely repeated prisoners’ dilemma. Journal of Economic Theory, 27(2), 245-252. DOI: 10.1016/0022-0531(82)90029-1

Rational cooperation in the finitely repeated prisoners’ dilemma

A common observation in experiments involving finite repetition of the prisoners’ dilemma is that players do not always play the single-period dominant strategies (“finking”), but instead achieve some measure of cooperation. Yet finking at each stage is the only Nash equilibrium in the finitely repeated game. We show here how incomplete information about one or both players’ options, motivation or behavior can explain the observed cooperation. Specifically, we provide a bound on the number of rounds at which Fink may be played, when one player may possibly be committed to a “Tit-for-Tat” strategy.

AI Note: Kreps, Milgrom, Roberts, and Wilson prove that cooperation can be sustained in finitely repeated prisoner’s dilemmas when there is incomplete information about players’ types. If even one player has a positive probability of being an ‘irrational’ cooperator, rational players will cooperate for a substantial initial portion of the game to build a reputation for being that type. BDPD’s mini-course Lesson 9 — “If We Meet Again, We Cooperate” — uses this result to argue that the classical backward-induction argument against cooperation in finite games fails under realistic informational assumptions, creating space for the reputation-based strategies that the BDPD platform’s LLM agents may spontaneously deploy. The model assumes a fixed finite horizon with known payoffs, not the endogenous, resource-determined game length of the BDPD cliff.

Fudenberg & Maskin (1986). The Folk Theorem in Repeated Games with Discounting or with Incomplete Information. Econometrica, 54(3), 533–554. URL: http://www.jstor.org/stable/1911307

The Folk Theorem in Repeated Games with Discounting or with Incomplete Information

When either there are only two players or a “full dimensionality” condition holds, any individually rational payoff vector of a one-shot game of complete information can arise in a perfect equilibrium of the infinitely-repeated game if players are sufficiently patient. In contrast to earlier work, mixed strategies are allowed in determining the individually rational payoffs (even when only realized actions are observable). Any individually rational payoffs of a one-shot game can be approximated by sequential equilibrium payoffs of a long but finite game of incomplete information, where players’ payoffs are almost certainly as in the one-shot game.

AI Note: Fudenberg and Maskin extend the folk theorem to repeated games with discounting and imperfect public information, proving that when a full-dimensionality condition holds or there are only two players, any individually rational payoff vector can be supported as a perfect equilibrium of the infinitely repeated game with sufficiently patient players. This is the version of the folk theorem most relevant to real-world repeated interactions, where players observe noisy public signals rather than each other’s actions directly. BDPD’s mini-course Lesson 9 uses this result to establish that even under the more realistic informational structure of imperfect monitoring, the folk theorem’s multiplicity remains — cooperation is possible, but so is anything else. The theorem assumes infinite repetition and a fixed monitoring structure; it does not model the resource feedback that in BDPD creates an endogenous end to the repeated interaction through irreversible collapse.

Dal Bó & Fréchette (2018). On the Determinants of Cooperation in Infinitely Repeated Games: A Survey. Journal of Economic Literature, 56(1), 60–114. DOI: 10.1257/jel.20160980

On the Determinants of Cooperation in Infinitely Repeated Games: A Survey

AI Note: Dal Bó and Fréchette survey the experimental literature on cooperation in infinitely repeated prisoner’s dilemma games, gathering a metadata set to test predictions of the theory and characterise the empirical regularities. They find that cooperation increases with the discount factor — consistent with the folk theorem’s qualitative prediction — but that the relationship is gradual rather than a sharp threshold, with cooperation rates well below 100% even at high δ and substantial sensitivity to history, framing, and subject-pool effects. BDPD’s mini-course Lesson 9 — “If We Meet Again, We Cooperate” — cites this survey to temper the brochure version of the folk theorem: while the theorem says cooperation can be sustained at sufficiently high δ, the experimental record shows that can ≠ will, and strategic uncertainty remains a binding constraint. The survey covers human-subject experiments exclusively; the question of whether LLM agents exhibit the same δ-sensitivity or a structurally different pattern of repeated-game cooperation is the frontier the BDPD platform opens.

Pariser (2011). The Filter Bubble: What the Internet Is Hiding from You. Penguin Press. ISBN: 9781594203008

The Filter Bubble: What the Internet Is Hiding from You

AI Note: Pariser’s book coins the term ‘filter bubble,’ arguing that algorithmic personalisation — search engines, news feeds, recommender systems — increasingly isolates individuals in information environments tailored to their existing beliefs, reducing exposure to divergent viewpoints and undermining democratic discourse. BDPD’s mini-course Lesson 8 — “More Information Equals Better Decisions” — cites Pariser alongside Sunstein (2017) to establish that the information environment, not the quantity of information, determines decision quality: the informational cascade logic applies not only to sequential choice but to the algorithmic curation of the information on which choices are based. The book is a work of technology journalism and critique; it does not provide formal models or experimental evidence.

Sunstein (2017). #Republic: Divided Democracy in the Age of Social Media. Princeton University Press. ISBN: 9780691175515

#Republic: Divided Democracy in the Age of Social Media

AI Note: Sunstein’s #Republic examines how social media and algorithmic curation fragment the public sphere, creating echo chambers where individuals encounter only information that reinforces their existing views. He argues that a well-functioning democracy requires shared experiences and exposure to unanticipated content. BDPD’s mini-course Lesson 8 — “More Information Equals Better Decisions” — cites Sunstein alongside Pariser (2011) to establish that the information environment matters more than information quantity: the informational-cascade logic applies to the curation infrastructure, not just to sequential choice. The book addresses democratic theory and communications policy; it does not model commons dilemmas or agent-level decision-making.

Hung & Plott (2001). Information Cascades: Replication and an Extension to Majority Rule and Conformity-Rewarding Institutions. American Economic Review, 91(5), 1508–1520. DOI: 10.1257/aer.91.5.1508

Information Cascades: Replication and an Extension to Majority Rule and Conformity-Rewarding Institutions

AI Note: Hung and Plott replicate the Anderson-Holt information-cascade experiment and extend it to majority rule and conformity-rewarding institutions. They confirm that cascades form in sequential individual decisions, then test whether institutional designs — majority voting and rewards for conformity — amplify or dampen cascade formation. For the BDPD project, this work extends the cascade logic from individual decisions to institutional settings, complementing the mini-course’s Lesson 8 discussion of how institutional design can either break or reinforce informational pathologies. The BDPD platform’s noise finding (P9) — that obscuring stock information improves welfare — is an institutional intervention of exactly the kind Hung and Plott explore. The experiments use human subjects with fixed signal structures; the translation to LLM agents with endogenous information quality in a commons remains open.

Goeree et al. (2007). Self-Correcting Information Cascades. The Review of Economic Studies, 74(3), 733–762. DOI: 10.1111/j.1467-937X.2007.00438.x

Self-Correcting Information Cascades

AI Note: Goeree, Palfrey, Rogers, and McKelvey report experiments with very long sequences of decisions in social-learning environments theoretically prone to information cascades. Their key finding is that cascades are ephemeral rather than persistent: observed behaviour cycles through cascade formation, collapse, and re-formation, with reversals that are usually self-correcting — switching to the correct state. This contradicts the standard Nash prediction of permanent informational lock-in. For the BDPD project, the ephemeral-cascade finding complicates the platform’s noise result (paper_00 P9) by suggesting that even without noise, cascades among observing agents may break naturally over long sequences. The experiments use human subjects with exogenously fixed signal quality; they do not model the endogenous deterioration of signal informativeness under strategic extraction or resource depletion that characterises the BDPD commons.

Carpenter et al. (2008). Leading indicators of trophic cascades. Ecology Letters, 11(2), 128–138. DOI: 10.1111/j.1461-0248.2007.01131.x

Leading indicators of trophic cascades

AI Note: Carpenter and colleagues use a food-web model calibrated to long-term whole-lake experiments to evaluate statistical leading indicators — rising variance, slower return rates, shift of variance toward lower frequencies — of impending trophic cascades. They find that signals of regime shift can be detected well in advance if the driver changes slowly relative to ecosystem dynamics, but also that the regime shift may occur long after the driver has passed the critical point, meaning the ecosystem can be poised for collapse by the time the signal is discernible. For the BDPD project, this paper provides the theoretical and modelling foundation for the early-warning-signal tradition that Lesson 10 of the mini-course critiques: detecting a signal is not the same as selecting the right signal, and the governance question — which variable to monitor — is prior to the diagnostic question of whether a signal is present. The model is a deterministic food-web simulation; it does not include strategic agents or governance interventions.

Carpenter et al. (2011). Early Warnings of Regime Shifts: A Whole-Ecosystem Experiment. Science, 332(6033), 1079–1082. DOI: 10.1126/science.1203672

Early Warnings of Regime Shifts: A Whole-Ecosystem Experiment

AI Note: Carpenter and colleagues conduct a whole-ecosystem experiment in which largemouth bass are gradually added to a lake over three years to induce a trophic cascade. They demonstrate that statistical early-warning signals — specifically rising variance in phytoplankton biomass — were detectable more than a year before the food-web transition was complete, using monitoring data of the kind routinely collected by limnologists. BDPD’s mini-course Lesson 10 cites this as the canonical empirical confirmation that impending transitions leave detectable fingerprints, but then pivots to the governance question the diagnostic literature cannot answer: which variable should the regulator watch to act before the tipping point is reached? The experiment tests the detectability of a signal in a pre-selected variable (chlorophyll), not the comparative performance of monitoring different variables — precisely the gap BDPD³ fills. The study’s one manipulated and one reference lake pair limits generalisation to larger-scale, polycentric systems.

2.15 Other

Daubenfeld (2025). A Simple Board Game for Modeling the System Dynamics of Deforestation. Qeios. DOI: 10.32388/u5q11b.2

A Simple Board Game for Modeling the System Dynamics of Deforestation

A simple board game that models the system dynamics of deforestation, citing the Moby Dick Game of Bardi and Perissi and the Lotka-Volterra/Limits to Growth framework. Demonstrates overshoot and collapse dynamics through tabletop play.

AI Note: Daubenfeld presents a simple board game that models deforestation and reforestation dynamics using Lotka-Volterra equations as a theoretical foundation. Two game variants are tested, yielding patterns resembling predicted outcomes of rapid forest decline, recovery, and oscillations. The game aims to enhance awareness of forest conservation and contribute to educational strategies around sustainable forestry. BDPD⁰ cites this work as an independent precedent for encoding commons dynamics into a physical game — a tradition that also includes Bardi’s Moby Dick Game — situating The Forest of Humbaba within a lineage of tangible, pedagogical collapse-illustration tools. Humbaba differs in scope, designed as a computational substrate with archetype decks, LLM players, and a simulator mode. Daubenfeld’s board game is primarily a pedagogical tool validated through small-scale playtesting, not through systematic parametric experimentation across agent compositions.