BDPD Research Notes
Six dispatches from the commons-laboratory: what the platform, the card game, and the agents have been telling us
About these notes
This volume collects the BDPD research notes — short essays written alongside the technical papers, the card game, and the simulation platform. Each note targets a single surprise: a result that came out differently from what we expected, an instrument that exposed a question the textbooks had not asked, a moment when the laboratory and the table-top game told us the same thing for the same reason. They are intended for readers who want to follow the project’s evolution without reading the formal papers first; the technical record is one click away, but the notes can be read on their own.
The six notes
1. What Happens When One Bad Actor Enters the Commons? — The first dispatch. A single aggressive agent collapses the collective survival rate to zero, regardless of pool size or regeneration rate. The threshold is discontinuous, the cliff edge is low, and the policy implications are direct.
2. Why Communication Wasn’t Enough — We gave our agents a voice. We expected cooperation to follow. We found that the cheap-talk literature’s strongest single intervention produced almost nothing when the speakers around the table were language models. The architectural channel matters more than the conversational one.
3. The Slow Catastrophe Has a Tell — On the Seneca effect and the leading/lagging signal distinction. Systems that grow slowly and collapse fast cannot be saved by monitoring the collapse signal. The variable you watch matters more than how well you watch it.
4. What a Card Game Knows That a Differential Equation Doesn’t — On The Forest of Humbaba, the table-top counterpart to the simulation. The card game produced the Tragedy of the Compensator finding — that cooperators subsidise defectors — before the equation did, and for a reason the equation could not capture on its own.
5. You Can’t Save the Polluter by Punishing Households — Three regulatory levers walk into a Seneca cascade. Two of them save the wrong thing. The lecture is in why.
6. The Mule, the Compensator, and the Inevitability of Late Defection — On graduated sanctions, late-stage defection, and why forty-five turns of cooperation provide zero resilience against a single strategic betrayal in poorly-designed institutions.
How to read
The notes are dated and can be read in any order. They overlap with — but do not summarise — the BDPD papers (Brunelli 2026a, 2026b, 2026c, 2026d) and with the ten lectures of the Limits of Good Will mini-course (Brunelli 2026e). The papers carry the formal record; the lectures carry the pedagogical arc; these notes carry the project’s temperature — what it felt like to discover each result, what we had been wrong about, and what came next. Because each note follows a single surprise rather than a single paper, the coverage is thematic rather than one-to-one: the governance-inversion result of BDPD² (Brunelli 2026c) — that the same institutions produce opposite outcomes for rule-bound and signal-adaptive agents — surfaces inside note 6 rather than in a note of its own.
The full project — code, paper sources, card-game artwork, and the documentation site — is available on GitLab. Code is AGPL-3.0; papers, course material, and game rules are CC BY 4.0; AI-generated card art is CC0.
Transparency and authorship. This document was authored by Roberto Brunelli with the support of frontier large language models; model outputs were source-grounded by prompt, validated by extensive triaging with different models and finally integrated under author supervision.