Adaptive coordination of working-memory and reinforcement learning in non-human primates performing a trial-and-error...

Area of Science:

• Cognitive Neuroscience
• Behavioral Economics
• Primate Cognition

Background:

• Human trial-and-error learning often involves a blend of reinforcement learning (RL) and working memory (WM).
• Comparing non-human primate neurophysiology to human brain activity is common, but it's unclear if monkeys use similar RL and WM processes.
• This study investigates the cognitive mechanisms underlying decision-making in monkeys during learning tasks.

Purpose of the Study:

• To determine if monkeys employ combined reinforcement learning (RL) and working memory (WM) strategies in discrete action tasks.
• To model and analyze monkey behavior using computational approaches that integrate RL and WM.
• To explore potential inter-species differences in the coordination dynamics between RL and WM.

Main Methods:

• Behavioral data from five monkeys performing trial-and-error learning tasks were analyzed.
• Computational models integrating reinforcement learning (RL) and working memory (WM) were used to fit trial-by-trial choices and reaction time data.
• Model-based analysis identified the best-fitting cognitive processes and coordination dynamics.

Main Results:

• Monkey behavior was best explained by a combination of reinforcement learning (RL) and working memory (WM), with significant inter-individual differences.
• The coordination dynamics between RL and WM varied across monkeys, with some showing patterns similar to previously studied humans, while others exhibited opposite dynamics.
• Analysis of model variants suggested that long pretraining might influence the coordination dynamics between RL and WM.

Conclusions:

• Monkeys, like humans, appear to use a combination of reinforcement learning (RL) and working memory (WM) for learning from discrete choices.
• Observed differences in RL-WM coordination dynamics may indicate inter-species variations or effects of experimental protocols.
• Further research in humans is warranted to explore these potential differences in cognitive process dynamics.