Kenza Kadri

PhD student | University of Plymouth

Linking transdiagnostic biomarkers to credit assignment impairment in Humans | Kenza Kadri

Linking transdiagnostic biomarkers to credit assignment impairment in Humans

Debriefing

Failure in decision-making can be related to some psychiatric traits such as compulsivity or anxiety. The overlap between psychiatric condition and poor decision-making can be illustrated with some examples, such as addiction or OCD. In such diseases, suffering individuals very often have difficulties predicting the consequence of their actions (Merikangas & McClair, 2012). Here, the impairment of decision-making can be due to the CA problem, where a patient assigned a less negative impact on drugs for example, than it really has. On the other hand, the CA problem could be due to a too quick response corresponding to habitual learning, that can be related to compulsivity. Currently, the link between CA problem and these psychiatric diseases has not been well established.

We suggest that we might be able to investigate the precise psychiatric phenotype associated with CA impairment and risk of developing a psychiatric disease, by building on the effectiveness of large-scale online data collection a general population, thus including statistically a large trait of psychiatric traits. In order to do so, we will use a computational-psychiatry approach based on mathematical models of decision making, to find the correlation between psychiatric traits and CA deficit (Beaumont et al., 2020). In particular, the purpose of this study is to investigate the relationship between the performances of a large cohort of healthy subjects on a decision-making task, with psychological traits collected using standardized questionnaires. The task aims to establish whether there is a difference in performance between individuals with different levels of compulsive behaviour, or any other psychological trait, and whether this interindividual difference is linked with failure in decisions estimated with Reinforcement Learning (RL) models. Thus, we hypothesize that a RL tailored to evaluate the CA problem will allow to highlight CA impairments and that, these impairments will be related to compulsivity behaviours.