An Empirically-Derived Taxonomy of Moral Concepts

We propose that methods from the study of category-based induction can be used to test the descriptive accuracy of theories of moral judgment. We had participants rate the likelihood that a person would engage in a variety of actions, given information about a previous behavior. From these likelihood ratings, we extracted a hierarchical, taxonomic model of how moral violations relate to each other (Study 1). We then tested the descriptive adequacy of this model against an alternative model inspired by Moral Foundations Theory, using classic tasks from induction research (Studies 2a and 2b), and using a measure of confirmation, which accounts for the baseline frequency of these violations (Study 3). Lastly, we conducted focused tests of combinations of violations where the models make differing predictions (Study 4). This research provides new insight into how people represent moral concepts, connecting classic methods from cognitive science with contemporary themes in moral psychology.