Figure 5.

Moral inference task. On every trial the agent chose between two options: more money for themselves plus more shocks for an anonymous receiver, or less money for themselves plus fewer shocks for the receiver. Participants predicted which option the agent would choose and subsequently received feedback about their accuracy. Every a few trials, participants provided their subjective impression of the agent's character on a scale ranging from “nasty” to “nice,” as well as their uncertainty of that impression.