Task Reliability Considerations in Computational Psychiatry

Abstract

Measuring psychological abilities or traits is trickier than it seems from the published literatures ( 1 ). We try to study abstract psychological constructs like ‘inhibition’ or ‘impulsivity’, but we can only measure these indirectly through behaviours such as reaction times or self-report ratings. When we apply these measures in clinical and individual differences research, our goal is typically to understand why a patient group appear to be (e.g.) more ‘impulsive’ than healthy controls, or to use measures of ‘inhibition’ to predict a clinical outcome. There are many potential pitfalls and wrong turns in the path towards achieving this goal. In our recent work, we have shown that our intuitions about what makes a good cognitive task can sometimes lead us astray ( 2 , 3 ). Here, we discuss how these issues intersect with the goals of computational psychiatry.