Title:	The cognitive and neural impact of perceived uncontrollability on reward learning
DNr:	NAISS 2025/22-336
Project Type:	NAISS Small Compute
Principal Investigator:	Marc Guitart-Masip <marc.guitart-masip@ki.se>
Affiliation:	Karolinska Institutet
Duration:	2025-03-03 – 2026-04-01
Classification:	30105
Keywords:

Abstract

Controllability—the belief that we can direct outcomes through our choices—shapes how we learn from experiences in complex environments. When our decisions yield desired results, a sense of controllability arises, reinforcing self-efficacy and leading to motivated goal-directed behaviour. However, when actions do not lead to desired consequences, a sense of uncontrollability arises, potentially leading to reflexive, passive behaviour and learned helplessness. Since stress and especially uncontrollable stress are known to contribute to some aspects of depression, learned helplessness is a popular model of anhedonia and depression. Moreover, it is known that stress, helplessness, and depression all cause different deficits in normal learning. Particularly, an inability to adapt to changing reward contingencies is a potential source of learning deficits in depression and anxiety. Therefore, a relevant gap in the literature is whether just a lack of control, an upstream preclinical factor, is enough to cause similar learning deficits in healthy humans. This research may offer novel targets for interventions aimed at stress-related disorders by pinpointing specific computational and neural mechanisms. Ultimately, these insights can potentially inform both basic neuroscience models of decision-making under uncertainty and clinical strategies that leverage a sense of controllability to foster resilience and improve mental health. By manipulating controllability in the laboratory, we aimed to identify the cognitive and neural consequences of perceived uncontrollability on flexible reward learning. In this preregistered fMRI study, 55 participants completed games of a probabilistic three-armed bandit task in an MRI scanner, where they chose between three composite stimuli comprising two images each. The goal was to choose the target, which changed about halfway through each game, introducing multiple orders of uncertainty. Additionally, participants aimed to avoid electric shocks or win monetary bonuses post-game. These outcomes were determined by their performance in controllable games but were random and unrelated to their performance in uncontrollable games. Goal outcome valence was manipulated orthogonally to controllability, such that participants experienced all combinations of conditions multiple times. Importantly, task difficulty, motivation, and success rates were matched across conditions. Secondarily, we also collected subjective ratings and anxiety (STAI) & depression (PHQ9) scales for each participant. The key questions that we ask are threefold: (1) Does perceived uncontrollability affect reward learning? (2) Does the effect of perceived uncontrollability differ when the outcome is shock avoidance versus monetary gain? (3) What are the neural representations of perceived controllability over shock avoidance, particularly in the orbitofrontal cortex and striatum? We will analyse behavioural, neuroimaging, and eye-tracking data to answer these questions. We will use various reinforcement learning and hidden Markov models to examine possible neurocomputational mechanisms underlying controllability effects on learning. Our behavioural analyses so far already show that subjective belief in controllability mediated the impact of controllability on learning after target reversal. We plan to do multiple univariate, multivariate, and functional connectivity analyses on the fMRI data to understand neural representations of controllability and outcome valence. The analysis plan, along with study details, is preregistered at https://osf.io/gav3e