Reward learning deficits are recognized as significant symptoms across various mental disorders. Recent studies suggest these learning impairments stem from a reduced capacity to utilize reward history to guide behavior. However, the neuro-computational mechanisms underlying these impairments remain largely unclear. Furthermore, there’s limited research employing a transdiagnostic approach to investigate whether the psychological and neural mechanisms contributing to learning deficits are consistent across different forms of psychopathology.
To shed light on this critical issue, we investigated probabilistic reward learning in individuals diagnosed with major depressive disorder (n = 33) or schizophrenia (n = 24), alongside a control group of 33 healthy participants. Our study combined computational modeling and single-trial EEG regression to analyze learning processes. In our task, participants were required to integrate the reward history associated with a stimulus to determine the worthiness of gambling on it. Adaptive learning within this task relies on dynamic learning rates, which are highest during initial encounters with a stimulus and decrease with repeated exposure. Ideally, choice preferences should stabilize over time, becoming less susceptible to misleading information as learning progresses.
Our findings reveal evidence of reduced Learning Dynamics in patient groups. Both the depression and schizophrenia groups exhibited hypersensitive learning, characterized by learning rates that decayed less significantly. This resulted in their choices being more vulnerable to misleading feedback. Additionally, we observed a schizophrenia-specific approach bias and a depression-specific heightened sensitivity to disconfirmational feedback, encompassing both factual losses and counterfactual wins. This inflexible learning observed in both patient groups was correlated with altered neural processing, notably the absence of expected value tracking in both patient groups.
Alt text: Illustration of brain activity during a learning process, highlighting neural networks involved in reward processing and decision-making, relevant to understanding learning dynamics in mental disorders.
Collectively, our results indicate that reduced trial-by-trial learning dynamics represent a shared deficit in both depression and schizophrenia. Moreover, we successfully identified distinct disorder-specific learning deficits, highlighting the complexity of reward learning impairments in psychopathology. Further research into learning dynamics will be crucial for developing targeted interventions for these debilitating conditions.
Alt text: Comparative graph illustrating learning rates across healthy controls, individuals with depression, and individuals with schizophrenia, emphasizing the reduced decay of learning rates in patient groups, indicative of altered learning dynamics.
References
