Introduction Social anxiety disorder or phobia (SAD) is a debilitating condition, where an individual experiences overwhelming fear to situations involving social interactions. Prototypically, SAD presents as shy, submissive, inhibited, and risk- aversive behaviours. Contrastingly, an atypical sub-group show impulsive, aggressive, novelty-seeking behaviours along with severe substance abuse problems. In scenarios, where there is co-existence of polar opposite symptoms, trans-diagnostic approaches extrapolate the characteristics of a disorder as a continuum rather than a categorical one. Data-driven computational models such as drift diffusion model utilize behavioural measures and extract potential markers that reflect the activity of specific brain networks. Here, we aim to analyse and correlate the psychological traits with computational estimates of behaviour during risk-taking and value based decision making.

Methods We used the data from 1400 participants who completed the 2 stage sequential learning task. We focused on the second stage of the task, where the reward probabilities of the choices are stochastic. The computational measures were estimated for two scenarios i.e. when the participants made 1) accurate choices and 2) risky choices (the choice with maximum variance in reward probability was labelled as risky). This computation was performed for all the trials across all the participants. We then used choice–(risky vs non-risky or correct vs incorrect) and response time as inputs to the hierarchical drift diffusion model to extract threshold (a), drift rate (v) and response bias (z) parameters. The computational parameters were then correlated with the 3 psychological factors that span the compulsive, anxiety- depression and the social withdrawal spectrum.

Results The computational parameters from both accuracy and risk taking scenarios of the sequential learning task were correlated with the 3 factors. While controlling for IQ and age, we found a generalized correlation which is significant between the threshold parameter(‘a’) and social withdrawal, with the former estimate being negatively correlated (Accuracy: |r| = -0.078, p=0.003; Risk: |r| = -0.075, p=0.005) with the latter. This relation was not observed with regard to anxiety-depression and compulsive traits.

Conclusions We show that individuals with higher social withdrawal levels are impulsive as they accumulate less evidence while making a choice. This behaviour holds irrespective of the choice being chosen is an optimal or a risky one. Critically, we show how a trans-diagnostic approach of integrating computational model and psychological questionnaires can reveal the existence of psychological traits as a continuum.