A newly tested model of social conformity, which discards outlier opinions rather than averaging them, has outperformed the traditional DeGroot model in five real-world experiments. Under the standard approach, individuals weigh all inputs equally, often yielding a compromise value.
The alternative model, previously published by Kaleda Denton and colleagues at the Santa Fe Institute, predicts that groups converge on the majority stance by ignoring dissenting views. In a turkey-roasting time example, six people suggest 15 minutes while one suggests 33; the new model matches the real outcome of 15 minutes, while averaging would produce 18.
Five distinct tests—covering cooking times, financial estimates, and other judgment tasks—found the new model more closely matched observed human behavior. The findings challenge the long-standing assumption that social influence operates through weighted averaging of all inputs.
If validated broadly, the model could reshape how researchers understand consensus formation in groups, from corporate boards to online communities. It also raises questions about how minority perspectives might be systematically lost in collective decision-making.
The authors caution that the experiments used relatively small groups and specific tasks, and further work is needed to test the model in larger, more diverse settings with higher stakes.