Model-Based Planning Deficits in Compulsivity Are Linked to Faulty Neural Representations of Task Structure

Abstract

Compulsive individuals have deficits in model-based planning, but the mechanisms that drive this have not been established. We examined two candidates—that compulsivity is linked to (1) an impaired model of the task environment and/or (2) an inability to engage cognitive control when making choices. To test this, 192 participants performed a two-step reinforcement learning task with concurrent EEG recordings, and we related the neural and behavioral data to their scores on a self-reported transdiagnostic dimension of compulsivity. To examine subjects' internal model of the task, we used established behavioral and neural responses to unexpected events [reaction time (RT) slowing, P300 wave, and parietal-occipital alpha band power] measured when an unexpected transition occurred. To assess cognitive control, we probed theta power at the time of initial choice. As expected, model-based planning was linked to greater behavioral (RT) and neural (alpha power, but not P300) sensitivity to rare transitions. Critically, the sensitivities of both RT and alpha to task structure were weaker in those high in compulsivity. This RT-compulsivity effect was tested and replicated in an independent pre-existing dataset ( N = 1413). We also found that mid-frontal theta power at the time of choice was reduced in highly compulsive individuals though its relation to model-based planning was less pronounced. These data suggest that model-based planning deficits in compulsive individuals may arise, at least in part, from having an impaired representation of the environment, specifically how actions lead to future states. SIGNIFICANCE STATEMENT Compulsivity is linked to poorer performance on tasks that require model-based planning, but it is unclear what precise mechanisms underlie this deficit. Do compulsive individuals fail to engage cognitive control at the time of choice? Or do they have difficulty in building and maintaining an accurate representation of their environment, the foundation needed to behave in a goal-directed manner? With reaction time and EEG measures in 192 individuals who performed a two-step decision-making task, we found that compulsive individuals are less sensitive to surprising action–state transitions, where they slow down less and show less alpha band suppression following a rare transition. These findings implicate failures in maintaining an accurate model of the world in model-based planning deficits in compulsivity.

Humans perseverate on punishment avoidance goals in multigoal reinforcement learning

“…Thus, we dissociated generalized from obsessive worry using the same approach taken in previous studies ( Doron et al, 2012 ; Stein et al, 2010 ), namely, by including a measure of OCD symptoms as a control covariate. Controlling for OCD symptoms has the additional benefit of accounting for known relations between OCD and poor learning of task structure, reduced MB control, and perseverative tendencies ( Gillan et al, 2016 ; Seow et al, 2021 ; Sharp et al, 2021 ). Secondly, another potentially confounding relationship exists between worry and somatic anxiety ( Sharp et al, 2015 ), likely reflecting a general anxiety factor.…”

A computational lens on the transdiagnostic nature of intolerance of uncertainty

“…These probabilities are referred to as state transitions, and participants' estimates of these state transitions can be inferred from the two-step task (Daw et al 2005). Recent applications of this task have revealed that individuals high in compulsive traits misestimate these probabilities, resulting in impaired learning (Seow et al 2021;Sharp et al 2021). Misestimating first-order uncertainty in the consequences of one's own action may then constitute a core aspect of compulsivity, and is thought to underlie checking behaviour in obsessive-compulsive disorder (Fradkin et al 2020).…”

“…Compulsive traits (Seow et al 2021;Sharp et al 2021) OCD (Marzuki et al 2022;Vaghi et al 2017) Adults with ASD (Arthur et al 2021)…”

Humans perseverate on punishment avoidance goals in multigoal reinforcement learning

“…Thus, we dissociated generalized from obsessive worry using the same approach taken in previous studies 62,63 , namely, by including a measure of OCD symptoms as a control covariate. Controlling for OCD symptoms has the additional benefit of accounting for known relations between OCD and poor learning of task structure, reduced model-based control, and perseverative tendencies [64][65][66] . Secondly, another potentially confounding relationship exists between worry and somatic anxiety 28 , likely reflecting a general anxiety factor.…”