Judging the intentionality of ambiguous action

Abstract

Judging the intentionality of others’ actions is a key aspect of social cognition; it gives meaning to actions and helps us predict what others will do. The ability to correctly judge intentionality is central to everyday social interactions as well as our justice system, where responsibility rests on the judgement of intentionality. Despite the clear importance of accurately judging other peoples’ action, it has been suggested that humans have a bias towards intentional attributions (Rosset, 2008). To explain this, Rosset (2008) introduced a dual-process model of intention attribution which suggests that there is an automatic tendency to judge all action as intentional, but that this can be overridden by higher-level controlled cognitive processes, leading to unintentional explanations of behaviour. Consequently, the model predicts that factors facilitating controlled processing (e.g., time to engage in processing, cognitive maturity linked to age, cognitive ability and availability of cognitive capacity) play a role in judging intentionality. In this thesis, I tested some of these predictions, however, apart from a replication of Rosset’s (2008) study, suggesting time pressure increases individuals’ intentionality endorsement, results do not support the dual-process model. Additionally, I investigated judging intentionality in Autism Spectrum Conditions (ASC) and the role of Theory of Mind (ToM). Individuals with ASC showed an increased tendency to attribute intent to ambiguous behaviour compared to neurotypicals, which could not be explained by differences in ToM. These results could indicate a difference in intention attribution style rather than failing to perceive mental states in ASC, which could help understand one aspect of social difficulties in ASC. In summary, the empirical evidence gathered in this thesis suggests that judging intentionality cannot be fully captured in a dual-process model. Therefore, at the end of this thesis, other approaches including a revised dual-process model and a single-system framework will be explored.