Action selection in early stages of psychosis: an active inference approach

Background and Hypothesis In order to interact successfully with our environment, we need to build a model, to make sense of noisy and ambiguous inputs. An inaccurate model, as suggested to be the case in psychosis, disturbs optimal action selection. Recent computational models, such as active inference (AI), have emphasized the importance of action selection, treating it as a key part of the inferential process. Based on an AI-framework, we examined prior knowledge and belief precision in an action-based task, given that alterations in these parameters have been linked to the development of psychotic symptoms. We further sought to determine whether task performance and modelling parameters would be suitable for classification of patients and controls. Study Design 23 at-risk-mental-state individuals, 26 first-episode psychosis patients and 31 controls completed a probabilistic Go/NoGo task in which action choice (Go/ NoGo) was dissociated from outcome valence (gain/ loss). We examined group differences in performance and AI-model parameters, and then performed receiver operating characteristic (ROC) analyses to assess group-classification. Study Results We found reduced overall performance in patients. AI-modelling revealed that patients showed increased forgetting, reduced confidence in policy selection and less optimal general choice behavior, with poorer action-state associations. Importantly, ROC-analysis revealed fair-to-good classification performances of all groups, when combining modelling parameters and performance measures. Conclusion Findingss: how that AI-modelling of this task not only provides further explanation for dysfunctional mechanisms underlying decision making in psychosis, but may also be highly relevant for future research on the development of biomarkers for early identification.