Active Inference of Interpersonal Trust Within and Across Mental Disorders

Background. Interpersonal problems are common in mental disorders like depression and social anxiety disorder. However, there is a lack of formal models to explain idiosyncrasies in patients' interpersonal functioning. Following modern computational accounts of perception and action, interpersonal experience and behavior results from implicit inference about hidden environmental properties.Methods. We simulate interpersonal decision-making in a “trust game” using a POMDP generative model and active inference. Simulated agents decide to either keep or invest an initial budget to a trustee. If the agents invests, the trustee can cooperate with or exploit the agent’s trust. Agents perform active inference of the context (cooperative vs. hostile) and action selection. By introducing systematic biases to the generative model, we simulate several subtypes of depression and social anxiety. We then collected data of N=25 patients with diagnosed depressive disorders and fit the generative model to the data. Results. Agents with biased expectations or preferences showed idiosyncratic interpersonal inference. They more readily infer to be in a hostile context and avoid investing with the trustee. Biases also affected the total earned rewards, with socially anxious and depressed agents receiving higher average rewards in hostile contexts compared to "healthy" agents (i.e., agents with an accurate-to-optimistic generative model), but “healthy” top optimistic agents showing superior performance in volatile environments. Fitting the model to the data resulted in individual parameters for patients with low symptom scores compared to those with high symptom scores, with group differences particularly in transition dynamics B, outcome preferences C and prior beliefs D. Discussion. Our simulations formalize complex interpersonal phenomena within Active Inference. Interpersonal biases in mental disorders can be formalized and investigated with respect to their functional role. The model has potential applications in psychopathological research, personalized diagnostics, and individualized treatment planning. Further empirical work is necessary to validate the empirical basis and clinical usefulness of formal models of interpersonal decision-making.