norm A norm B (the shape is the coordination norm)
How the model works
Each generation every dot meets two partners, from its own group unless it reaches out. Cooperating costs c and gives the partner b; sharing a norm gives d. Then a share of dots learn: most copy someone in their group, more often when that someone earned more; some conform to their group's majority; a few learn from other groups. Now and then a richer group converts a poorer one.
CFst is the share of the variation in cooperation that lies between groups. When partners come from inside the group, it sets how far a dot's cooperation predicts its partner's, which is β on the slide. Cooperation is favoured while β stays above c/b. Coordination needs no such help, since matching the majority already pays, so norm clusters form fast. The slide comes from a talk by Joseph Henrich. Cultural Fst is from Bell, Richerson and McElreath (PNAS, 2009), who found far more variation between neighbouring groups in culture than in genes, which leaves much more room for selection between groups. The assortment rule itself is Fletcher and Doebeli's (2009).