Philadelphia’s Got Talent! A novel, naturalistic paradigm for inducing socio-affective prediction errors

Prediction error — the discrepancy between what is expected and what actually occurs — is increasingly recognized as a core computational principle organizing perception, action, and affect. Embarrassment can be conceptualized as a socio-affective prediction error, arising when social outcomes violate expectations. Here, we introduce a task to systematically induce embarrassment and measure these prediction errors, called Philadelphia's Got Talent (PGT). In the first session, participants sang, danced, or told a silly story while being videotaped. A week later, they unexpectedly viewed footage of their own and another participant's performance while providing continuous ratings. Results show that participants underestimated how positively others would evaluate their performances — a "liking gap" — and exhibited socio-affective prediction errors specifically during norm-violating performances, but not during control conditions. Participants self-reported feeling embarrassed by their own performances but amused by other people’s performance. Together, these findings support PGT as a paradigm for eliciting0 and quantifying socio-affective prediction errors, capturing individual differences in responses to social norm violations, and offering a computational framework for future research in social psychology and social neuroscience as well as clinical science.