Does it matter how far back a fading ally looks: the last partner or the whole memory?

What it found

The society-allywatch crowd (200 people, memory 9, error 0.02, 20% allies, 12 seeds, 1500 rounds) with allies who convert at rate c×f, f the share of their last k partners who walked the other way, for k = 1, 3, 5 and 9 (9 reproduces society-allywatch exactly). At c = 0.01 and robot share 0.10 all look-backs tip 11–12 of 12 crowds against 10 for constant fading; at c = 0.03 and 0.15 robots all tip 12 of 12 against 10. In the marginal cell (c = 0.03, robots 0.10) it is 2, 3, 3, 4 of 12 against 0, a monotone but one-crowd-per-step trend that 12 seeds cannot separate. Stylised: the rule and c are assumed, not measured; simulated walkers only.

A figure from its own output

Paper, code and results

• README and method
• Experiments and results
• Code
• Tests

More in Robot societies among people

• If allies fade when they see the crowd walking away from them, does the robots' tip survive?Allies who convert only when they have watched the crowd walk the other way. With 20% allies, fading at rate c times the share of R in their memory tips more crowds than a constant rate c once robots are 0.10 or more (12/12 vs 10/12 at c = 0.01, robots 0.10; 4/12 vs 0/12 at c = 0.03), but does nothing below 0.05. Stylised
• How much cheaper must an urgency cue be for the urgent before a door robot can rely on it?How much cheaper must an urgency cue be for the urgent before a door robot can rely on it? Urgent people must pay at most about half what normal people do; the crowd is worse off than under yield-all in every cell.
• At what collision cost does a door stop tempting robots to go first?At what collision cost does a door stop tempting robots to go first? With 30% robots and dithering cost 1, always-yielding robots are individually stable only above a collision cost of 2.80 (waiting cost 1); below it a single robot gains by going though yielding is best for the crowd. The threshold rises with dithering cost and robot share.
• When robots always yield at a door, do people learn to walk through first, and who pays?When robots always yield at a door, do people learn to walk through first, and who pays? Always-yielding robots raise the share of people who go from 0.388 to 0.595 at 50% robots and cut the crowd's cost (1.064 to 0.968); always-going robots raise it to 1.373. Mean field and simulation agree within 0.003.
• Can a few queue-enforcing robots hold a queueing norm against line-cutting robots?How many enforcer robots hold a queue against line-cutting robots? About 5% of the crowd holds the norm against 30% cutters (memory 9, 10 people in view of each cut, stop chance 0.5), simulation agrees. Stylised: view size and stop chance are swept, not sourced.
• Do stubborn walkers who convert after meeting robots let a smaller robot minority tip a crowd's walking side?Do stubborn walkers who convert after meeting robots let a smaller robot minority tip a crowd's walking side? At 24% robots every crowd tips for k = 30 meetings (median round 171) though the fixed block stops it; at 16-20% none tip.

All 31 Robot societies among people projects →

Other research topics

Proper scoring & elicitation28Effort & peer prediction9Verification games & slashing19Detection & tolerances14Markets as routers13Prediction-market design9Paying contributors8Decentralised training systems23Digital twins & sim-to-real164LLM chip design loops40Robot manufacturing loops53Ambient AI assistant46Self-replicating manufacturing34Simulation-first design64Engineered clothing45Task-specific AI accelerators40Algorithmic chip-design loops42Open-source purpose-built robots44The software-engineering loop43Adversarial human imitation41Open-source residential construction41Autonomous multi-angle capture34Sensing, imaging and 3D31Energy supply and dense energy storage39Robot societies among people31Agentic co-design of low-cost robots27Laundry-hanging robots25Machine learning on qualitative data23Robots on the building site27Civilizations & group psychology1History & human sciences1

Produced with AI assistance (Claude). Results are from the code in this repository and are simulated or stylised unless the abstract says otherwise; nothing here implies affiliation with or endorsement by any lab, researcher or company cited. MIT licensed.