Anxious and obsessive-compulsive traits are independently associated with valuation of noninstrumental information

Aversion to uncertainty about the future has been proposed as a transdiagnostic trait underlying psychiatric diagnoses including obsessive-compulsive disorder and generalised anxiety. This association might explain the frequency of pathological information-seeking behaviours such as compulsive checking and reassurance-seeking in these disorders. Here we tested the behavioural predictions of this model using a non-instrumental information-seeking task that measured preferences for unusable information about future outcomes in different payout domains (gain, loss, and mixed gain/loss). We administered this task, along with a targeted battery of self-report questionnaires, to a general-population sample of 146 adult participants. Using computational cognitive modelling of choices to test competing theories of information valuation, we found evidence for a model in which preferences for costless and costly information about future outcomes were independent, and in which information preference was modulated by both outcome mean and outcome variance. Critically, we also found positive associations between a model parameter controlling preference for costly information and individual differences in latent traits of both anxiety and obsessive-compulsion. These associations were invariant across different payout domains, providing evidence that individuals high in obsessive-compulsive and anxious traits show a generalised increase in willingness-to-pay for unusable information about uncertain future outcomes, even though this behaviour reduces their expected future reward. Author Note Daniel Bennett, Princeton Neuroscience Institute, Princeton University and Melbourne School of Psychological Sciences, The University of Melbourne; Kiran Sutcliffe, Nicholas Poh-Jie Tan, and Luke D. Smillie, Melbourne School of Psychological Sciences, The University of Melbourne; Stefan Bode, Melbourne School of Psychological Sciences, The University of Melbourne and Department of Psychology, University of Cologne. This work was supported by an ARC Discovery Project grant (DP180102383) to SB. DB receives salary support from an NHMRC CJ Martin Early Career Fellowship (1165010). We thank Trevor Chong and Sam Zorowitz for comments on an earlier draft. De-identified behavioural and self-report data, as well as analysis code (for reproducing correlational analyses and Stan models) are publicly available in an Open Science Framework repository at https://osf.io/eg74d/ .