The Role of Working Memory in Reward-based Learning over Long Timeframes: Opponent Bootstrapping and Interference E ects of Working Memory on Incremental Learning

Learning from feedback is essential for effective decision-making in daily life. Previousexperiments have indicated that feedback-based learning involves multiple processes, suchas slower reinforcement learning (RL) and fast working memory (WM). However, most ofthese studies focused on the dynamics of RL and WM in short timeframes, ranging from afew seconds to minutes, whereas real-life learning may occur over hours, days, or weeks.This preregistered study aimed to investigate the interplay between RL and WM inlong-term feedback-based learning. Instead of learning back-to-back iterations in quicksuccession, participants learned one or two iterations per day for 24 days. In this scenario,WM should only assist participants in conditions where a given item appears more thanonce in short succession. The results show that (1) when WM was unavailable, the size ofthe learning sets did not significantly influence learning performance; (2) when WM wasavailable, it had both immediate and long-term effects in boosting learning; and (3) thedeleterious interference effect of WM on long-term retention was observed during learning,but overshadowed in the test phase and modeling results by its ability to lead to moreexperience of rewards, illustrating the complexity of interactions between multipleprocesses supporting learning.