Reciprocity and reputation
Can a small concession buy enough future cooperation to recover its cost?
01Make a concession
02Observe reciprocity
03Update trust carefully
The decision at the table
You can accept a slightly worse trade now because you expect better access to trades later. That may be sensible even for a self-interested player. It may also just subsidize an opponent who will never reciprocate.
The mechanism to test
Model future response conditional on the specific partner, the observed interaction, and the remaining horizon. Distinguish generosity, reliability, reciprocity, and perceived threat. Keep separate beliefs for these dimensions. Start with within-game memory; cross-game reputation needs an explicit persistent-identity and memory condition.
Proposed experiment
Compare no social memory, recipient-specific memory with no concessions, and bounded concessions with that same memory. Hypothesis: concessions recover their cost against reciprocal partners and fail against nonreciprocal partners. Add forgiveness versus permanent punishment as a follow-up, not a hidden change.
Use an event-aligned trajectory showing the concession, later offers, acceptance, and cumulative foregone reference value. Log opportunities to reciprocate as well as actual reciprocation. Whole-game randomization estimates the policy effect; the later accepted-trade subset cannot identify causality on its own.
First ablation
Scaling modeled willingness by each partner's observed acceptance rate changed the contrast by -0.031 wins per game (95% interval -0.125 to +0.063) on 64 deterministic boards (90 treatment wins against 98): against builders that accept by a fixed deficit rule, learning who accepts does not pay. A reciprocal opponent is the missing ingredient, and the fixed builders cannot supply it. Run 7c17e2c7-a140-4fe5-b055-68bdea33ea60; see the ablation report.
What could disprove it
Seat numbers identify people only within a game. A fresh game must not silently inherit a belief about “seat 2.” Cross-game evaluation clusters by persistent group or session, and fixed known horizons may weaken repayment incentives.
Agent notes
Use the shared experiment design to freeze candidate versions, full lineups, budgets, sample size, primary contrast, and stopping rules before collecting evidence. This is a draft study brief, not a preregistration. No run IDs exist for this proposal.
The deployable policy reads only its own observation and recipient-visible events. The diagram is a conceptual schematic. Build new measured exhibits from retained artifacts using the visual publishing guide.
Try the pliability laboratory to see an explicit toy model of when a concession becomes eligible.
All approaches · Player’s guide · Experiment program
Tracked investigations
| Updated | Investigation | Status | Finding and next step |
|---|---|---|---|
| 2026-09-09 | Learning who accepts | Active | Round one, 256 deterministic games: contrast -0.031 (interval -0.125 to +0.063); no effect against fixed builders. Next: Test reciprocity against an opponent whose acceptance depends on history, for example a second search seat with the same model. Log 2026-09-09 |