Computational Phenotyping: Using Models to Understand Individual Differences in Personality, Development, and Mental Illness

Abstract

This paper reviews progress in the application of computational models to personality, developmental, and clinical neuroscience. We first describe the concept of a computational phenotype, a collection of parameters derived from computational models fit to behavioral and neural data. This approach represents individuals as points in a continuous parameter space, complementing traditional trait and symptom measures. One key advantage of this representation is that it is mechanistic: The parameters have interpretations in terms of cognitive processes, which can be translated into quantitative predictions about future behavior and brain activity. We illustrate with several examples how this approach has led to new scientific insights into individual differences, developmental trajectories, and psychopathology. We then survey some of the challenges that lay ahead.

Interfacing neural constructs with the Hierarchical Taxonomy of Psychopathology: ‘Why' and ‘how'

“…trait neuroticism) and neural responses have been small and inconsistent because of the substantial theoretical gap between what is measured via these modalities. They advocated for computational modelling as a promising technique for bridging these units of analysis through modelling of latent intermediate processes and testing theories of how and why separable processes lead to measurable behaviours, symptoms, and neural responses . Computational modelling has also been used to elucidate the mechanisms of cognitive control deficits in schizophrenia .…”

“…Another important feature of computational modelling is its applicability to research on individual differences. Estimates can be derived from a given model in the form of ‘computational phenotypes' or ‘computational multidimensional profiles' that reflect between‐subjects variation in the mechanisms of interest. In this way, an individual's performance on a well‐designed task could be used to compute an estimate of dysfunction in a particular cognitive process and then linked to a dimension of psychopathology.…”

“…Estimates can be derived from a given model in the form of ‘computational phenotypes' or ‘computational multidimensional profiles' that reflect between‐subjects variation in the mechanisms of interest. In this way, an individual's performance on a well‐designed task could be used to compute an estimate of dysfunction in a particular cognitive process and then linked to a dimension of psychopathology. For example, Sevgi et al demonstrated that individual computational phenotypes reflecting reliance on social information during a decision‐making task were associated with scores on a measure of the autism spectrum.…”

“…For example, Sevgi et al demonstrated that individual computational phenotypes reflecting reliance on social information during a decision‐making task were associated with scores on a measure of the autism spectrum. Further, including self‐report data in a computational model along with behavioural variables appears to improve its predictive accuracy …”

RDoC at 10: changing the discourse for psychopathology

“…Such developments can advance multiple aspects of the needs for precision diagnosis. Experimental paradigms developed with computational models are showing the potential to delineate specific aspects of behavior, and to relate these various aspects to their implementation, coordinated by increasingly well‐specified brain circuits 10 . This development capitalizes on the RDoC principle to explain relations across units of analysis in order to clarify psychophysiological constructs and the critical relationships among interrelating response systems.…”