Identification of transdiagnostic computational biomarkers of the brain for social interaction disorders using neuroimaging hyperscanning

点击翻译按钮获取中文摘要

Actions and decisions require learning about the environment and outcomes of our choices, which are often uncertain. In a social context, we learn about outcomes that depend on one’s own, but also another’s actions. This is complicated through the reciprocity of interaction (each decision affects the partner’s future decisions, which later affects me, etc.), and because information is not directly accessible, but has to be inferred from observations (e.g., a smiling person is friendly). Paradox decision making (we act, when it is irrational, e.g., playing in the lottery; or do not act, although probabilities would suggest so, e.g., eating less food to reduce obesity) is relevant in mental health, as causal disturbances are transdiagnostic features of mental illnesses, and imply a shift in decision making to the disadvantage of the patient (e.g., perceived loss of control during depressive episodes).In my fellowship, I aim to identify ecologically valid social interaction-based biomarkers through computational modelling to gain mechanistic insights into the biological underpinnings of social interaction disorders.I propose a renewal of my fellowship to allow me to extend my expertise to interpersonal learning, a core contributor to social interaction disorders. To achieve this, I will collaborate with Professor Robin Murphy at the Department of Experimental Psychology at Oxford University. We will present a novel interpersonal learning task, derived from his extensive studies on learning and decision making. In this task, subjects learn action-outcome contingencies and perceived certainties, while outcomes depend on the actions of both subjects.This data will be included in my efforts on computational modelling of social interaction, so that I will be able to highlight the key parameters that lead to the impaired social learning. Underlying causes may rest upon a failure to attend to social information, or aberrant beliefs about the precision or importance of social cues. These result in a failure to update prior knowledge. When identified by model parameters capturing elements of the social inference process, we can use these markers as a target for intervention that is tailored to the individual needs of a patient, independent of diagnostic categories. Consequently, I will generate a fully inclusive computational model of social interaction dysfunction.In addition, I propose a visit to Professor Read Montague, heading the Computational Psychiatry Unit at the Virginia Polytechnic Institute and State University. Here I will apply my multi-agent models of active inference to a full sample of opmMEG data, which allows to examine social functioning at the timescale it occurs.This work implies a significant advancement of my work, and provides in-depth expertise on opmMEG data and its computational modelling. Our joint data set on economic exchange paradigms on healthy and clinical subjects will represent an unmatched foundation for modelling.