Empathic forecasting in a cancel culture: how well do we predict what offends others?

Across three studies (256 participants making 6,866 scenario-level ratings), we investigate how people forecast others' emotional responses to potentially offensive behaviours. Our two goals were to assess whether these forecasts reflect social projection, i.e. whether people rely on their own emotional reactions when estimating others', and to assess the accuracy of those forecasts using both intensity ratings and percentage estimates (sometimes called consensus estimates). In Study 1, participants rated how offensive 15 behaviours felt to them and how offensive they believed others would feel. Self and other ratings were strongly correlated, but participants consistently believed others would be more offended, especially for behaviours most people regarded as benign. In Study 2, participants indicated (yes/no) whether they were offended by 32 behaviours and estimated what percentage of others would be offended. We found evidence of a truly false consensus effect; those who were offended showed substantial overestimation, although even unoffended participants overestimate percentages for the least offensive behaviours. Study 3 replicated these patterns using a hybrid design. We discuss possible mechanisms underlying our findings, as well as implications for the regulation of offensive content and directions for future research.