Policy shaping based on the learned preferences of others accounts for risky decision-making under social observation.

Observing others' choices influences individuals' decisions, often leading them to follow others. However, it is repeatedly reported that being observed by others tends to make people act more riskily. We hypothesized that this discrepancy arises from individuals' belief that others prefer riskier choices than they do. To examine this hypothesis, we used a gambling task where on some trials, individuals were informed that their choices would be observed by a risk-averse or seeking partner. Most important, individuals were given chances to learn each partner's preference beforehand. As expected, individuals initially believed that partners would make relatively riskier choices than they would. Against two alternative explanations, we found that individuals simulated partners' choices and weighed these simulated choices in making their own choices. Using functional magnetic resonance imaging, we showed that decision probabilities adjusted with the simulated partners' choices were represented in the temporoparietal junction (TPJ). Moreover, individual differences in the functional connectivity between the TPJ and the dorsomedial prefrontal cortex (dmPFC) were explained by the interaction between model-estimated social reliance and sensitivity to social cues in the dmPFC. These findings provide a neuromechanistic account of how being observed by others affects individuals' decision-making, highlighting the roles of the dmPFC and TPJ in simulating social contexts based on individuals' beliefs.