Linking reinforcement learning, working memory, and choice dynamics to age and symptoms of anxiety and depression in adolescence

Adolescence is a sensitive period characterized by significant neurocognitive development, with important implications for learning and decision-making. Working memory and reinforcement learning are essential for decision making and real-life dynamic adaptation to the environment and are often affected in individuals with anxiety and depression. Using a cognitive computational approach, we investigated associations between working memory, reinforcement learning, age, and symptoms of generalized anxiety and depression in 193 adolescents aged 12-24 years. Participants completed the Reinforcement Learning Working Memory (RLWM) task. To gain insight into the dynamics underlying instrumental learning behavior, we employed a computational model that combines the RLWM model with a Linear Ballistic Accumulator (RLWM-LBA), to quantify processes related to working memory and reinforcement learning, as well as choice dynamics underlying reaction times. We observed an age-related increase in task performance, but no differences depending on symptoms. Bayesian regression models revealed strong evidence for an association between age and the start point variability parameter of the LBA module, suggesting reduced choice stochasticity in older adolescents. In line with the behavioral findings, we found anecdotal to moderate evidence for no associations between RLWM-LBA parameters and symptom sum scores. Lastly, we trained regression models testing the utility of RLWM-LBA parameters for predicting age and symptom burden, yielding poor predictive performance. This study highlights differences in choice dynamics underlying age-related improvements in instrumental learning, while finding that cognitive differences identified in case-control studies did not generalize to symptom variation in this sample from the general adolescent population.