Disc tint: share cooperating SYNTHETIC
Ring weight: workers who work at home OPEN
Levers
Most self-contained towns OPEN
Most connected towns OPEN
How the model works, and credits
Each town is a group of dots, sized by the square root of its population. Every generation each dot meets two partners. A dot from town h meets someone in town w with probability equal to the share of h's commuters who work in w (LEHD LODES 2023), times the everyday mixing setting; every other meeting stays in town. Cooperating costs c and gives the partner b; sharing a coordination norm gives d. Dots then learn: they copy a partner with a probability that rises with the partner's payoff, lean toward their town's majority, and sometimes learn from another town, chosen by the same commuting weights. Richer towns can convert poorer neighbours. CFst is the share of variation in cooperation that lies between towns; β is measured from the meetings as P(partner cooperates | cooperator) minus P(partner cooperates | defector).
Levers: a bridge adds mixing between two towns; an anchor institution makes meetings inside a town assortative (cooperators find cooperators) without cutting contact; a showcase town is copied more by the others; a norm campaign raises conformity in one town. Drive minutes are OSRM centroid to centroid (derived), border waits excluded. Commuting counts measure contact, and the model's cooperation is a simulated trait: it describes no real town.
Credits: Joseph Henrich's talk on cultural evolution and assortment; Bell, Richerson and McElreath, "Culture rather than genes provides greater scope for the evolution of large-scale human prosociality", PNAS (2009); Handley and Mathew, "Human large-scale cooperation as a product of competition between cultural groups", Nature Communications (2020); Muthukrishna et al., "Beyond Western, Educated, Industrial, Rich, and Democratic (WEIRD) psychology", Psychological Science (2020). Data: Census places 2023, ACS 2023 B01003, LEHD LODES8 WA 2023, OSRM.