Different Algorithmic Models Underlie Virtue and Vice Attributions

Abstract

One of the most pressing questions in social psychology is how people update character attributions about other people considering novel information. A possible way to tackle this question is to algorithmically model trait attribution updating and confront it to how people actually update character attributions. Here, we present, parameterize, and empirically test several Bayesian and averaging models of character-based moral judgment over multiple pieces of morally relevant or distractor information. Taken as a whole, results from two experiments suggest that virtue and vice attributions follow different algorithms. Depending on the structure of received information virtue and vice attributions can follow differently weighted Bayesian algorithms or average-based models. We discuss these results in light of both classic findings in moral psychology and cognitive sciences in general.