Fig. 6: Experiment 3 Methods and Results.

a We reared two groups (orange and blue fish) separately in small white virtual cups, akin to the rearing conditions from the biological fish study (Engeszer et al.⁵⁰). After training, we froze learning in the artificial fish and tested their social preferences with the 2AFC task used in Engeszer et al.⁵⁰, measuring whether each fish spent more time with the familiar versus novel group. The red, green, and blue coordinate lines are for visualization and were not visible to the artificial fish. b Results from the 2AFC task. Each bar shows the social preference score (proportion of time spent with familiar versus novel group) of a single fish (n = 8). The fish reared in the orange group spent more time with orange fish versus blue fish, and the fish reared in the blue group spent more time with blue fish versus orange fish. Most of the artificial fish spontaneously developed social preferences for in-group members. c To measure the impact of social experiences in learning social preferences, we reared artificial fish in non-social environments, without other artificial fish. During training, all fish acquired experience with the visual environment, but the fish reared separately did not acquire experience with social partners. Compared to the artificial fish reared in groups, the artificial fish reared separately showed little to no evidence of social grouping in the 2AFC task. The development of social preferences in the artificial fish required experience with social partners, akin to the development of social preferences in newborn fish, who develop preferences for social partners encountered early in life. One-sample t-tests were performed to determine statistical significance (*p < 0.05, **p < 0.01, and ***p < 0.001, uncorrected; the exact p values and raw values are provided in the Source Data). Data are presented as mean values. Error bars indicate ±1 SEM.