Assigning the right credit to the wrong action: compulsivity in the general population is associated with augmented outcome-irrelevant value-based learning

Abstract

Compulsive behavior is enacted under a belief that a specific act controls the likelihood of an undesired future event. Compulsive behaviors are widespread in the general population despite having no causal relationship with events they aspire to influence. In the current study, we tested whether there is an increased tendency to assign value to aspects of a task that do not predict an outcome (i.e., outcome-irrelevant learning) among individuals with compulsive tendencies. We studied 514 healthy individuals who completed self-report compulsivity, anxiety, depression, and schizotypal measurements, and a well-established reinforcement-learning task (i.e., the two-step task). As expected, we found a positive relationship between compulsivity and outcome-irrelevant learning. Specifically, individuals who reported having stronger compulsive tendencies (e.g., washing, checking, grooming) also tended to assign value to response keys and stimuli locations that did not predict an outcome. Controlling for overall goal-directed abilities and the co-occurrence of anxious, depressive, or schizotypal tendencies did not impact these associations. These findings indicate that outcome-irrelevant learning processes may contribute to the expression of compulsivity in a general population setting. We highlight the need for future research on the formation of non-veridical action−outcome associations as a factor related to the occurrence and maintenance of compulsive behavior.

Cognitive resources predict value learning for outcome-irrelevant features

“…We found strong evidence for outcome-irrelevant learning, such that participants were more likely to stay with their response key selection after the first offer was rewarded (48%) vs. unrewarded (45%; posterior median =0.09, HDI95% between 0.03 and 0.16; probability of direction (pd) 99.75%; 0% in ROPE (-0.013 -0.013) and Bayes Factor (BF) of 7.66 against the null and of 7.29 against the modified ROPE; Figure 2C). This replicates previous findings of outcome-irrelevant learning in human individuals (Rmus, Zou, et al, 2021;Shahar et al, 2019Shahar et al, , 2021. Importantly, we found support for our pre-registered hypothesis that individuals with low working memory capacity will demonstrate increased outcome-irrelevant learning compared to high capacity individuals such that the interaction between previous-outcome and capacity was negative (posterior median =-0.14, HDI95% between -0.22 and -0.07; probability of direction (pd) 100%; 0% in ROPE (-0.013 -0.013) and Bayes Factor (BF) of 280 against the null and of 289 against the modified ROPE; Figure 2E).…”

“…However, little is known regarding how cognitive control processes might prevent individuals from assigning credit to aspects of the environment that are knownwith high certaintyto be irrelevant to the observed outcome (Rmus, McDougle, et al, 2021). To the best of our knowledge, this preregistered study, was the first to demonstrate a moderating role for a well-known cognitive control system, namely working memory capacity (Cowan, 2017), on the tendency to assign value to irrelevant features of the environment (Rmus, Zou, et al, 2021;Shahar et al, 2019Shahar et al, , 2021.…”

“…Importantly, only the cards predicted reward, but not locations or response-keys used to report cards' selection. This fact is important since it renders any credit assignment to the location/response key as outcome-irrelevant learning (Shahar et al, 2019(Shahar et al, , 2021. Participants were told explicitly during the instruction phase that only cards predicted reward and not the response-keys used for their selection (see SI).…”

“…£0 vs £1) affected response key selection in the second offer (see Figure 2A). Specifically, we reasoned that credit will be assigned not only to the chosen card but also to a response key used for its selection (Shahar et al, 2019(Shahar et al, , 2021. This would mean that after reward (i.e., £1) was obtained in the first offer, participants should be more likely to stay and choose the second offer with the same response key that was selected in the first one.…”

“…The visual and tactile cues of each fruit should predict its taste, whereas its position on a table and the hand with which it is taken should be deemed irrelevant and thus not assigned value nor be considered in the child's choice of fruit. Recent studies have indicated that individuals attribute credit to outcome-irrelevant features of an action, even when they hold certain and explicit knowledge that those features have no predictive value for the outcome, a phenomenon we coined "outcome-irrelevant learning" (Rmus, Zou, et al, 2021;Shahar et al, 2019Shahar et al, , 2021. Studying outcome-irrelevant learning should shed light on the manner in which people assign credit to various features during reinforcement learning, yet the underlying processes that moderate outcome-irrelevant information during credit assignment are still largely unknown.…”