Individuals with anxiety and depression use atypical decision strategies in an uncertain world

Abstract

The theory of optimal learning proposes that an agent should increase or decrease the learning rate in environments where reward conditions are relatively volatile or stable, respectively. Deficits in such flexible learning rate adjustment have been shown to be associated with several psychiatric disorders. However, this flexible learning rate (FLR) account attributes all behavioral differences across volatility contexts solely to differences in learning rate. Here, we propose instead that different learning behaviors across volatility contexts arise from the mixed use of multiple decision strategies. Accordingly, we develop a hybrid mixture-of-strategy (MOS) model that incorporates the optimal strategy, which maximizes expected utility but is computationally expensive, and two additional heuristic strategies, which merely emphasize reward magnitude or repeated decisions but are computationally simpler. We tested our model on a dataset in which 54 healthy controls and 32 individuals with anxiety and depression performed a probabilistic reversal learning task with varying volatility conditions. Our MOS model outperforms several previous FLR models. Parameter analyses suggest that individuals with anxiety and depression prefer suboptimal heuristics over the optimal strategy. The relative strength of these two strategies also predicts individual variation in symptom severity. These findings underscore the importance of considering mixed strategy use in human learning and decision making and suggest atypical strategy preference as a potential mechanism for learning deficits in psychiatric disorders.