Modeling Responsibility Attribution in a Group

Attributing responsibilities to a group of agents who contribute to a collective causeis common in our daily life. While we as humans could relatively efficiently assignresponsibility to others using our intuition, the wide range of possible scenarios andcomplexity of our cognitive mechanisms should not be overlooked. Understandingof how rational responsibility judgments are made could expand our knowledgeabout human causal and moral reasoning. Over recent years, a few theories havebeen developed to account for responsibility attribution, but they have various theoreticaland empirical limitations. In this thesis, an extended computational modelis proposed in an effort to expand the explanatory power of previous models. Wedevised two sets of experiments to test the performance of this newmodel, especiallywith regards to aspects including (1) epistemic state of rater, (2) perceived autonomyof agents, and (3) value range of contributions. Our results suggest that the newmodel demonstrated reasonable power in predicting responsibility attribution, albeitwith several unexpected inaccuracies. The close relationship among responsibility,criticality, and pivotality is nonetheless supported.