Exploring strategy differences between humans and monkeys with recurrent neural networks.

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Artificial Intelligence

Background:

• Non-human primates are key models for decision-making research in cognitive neuroscience.
• Discrepancies exist between monkey and human cognition, raising concerns about extrapolating findings.
• Working memory tasks reveal monkeys use recency strategies, while humans use target-selective strategies.

Purpose of the Study:

• To investigate the cognitive discrepancy between monkeys and humans using artificial neural networks (ANNs).
• To explore the learning progression in ANNs on a working memory task.
• To understand the underlying mechanisms of apparent recency-based strategies.

Main Methods:

• Utilized artificial neural networks (ANNs) as a parallel model system.
• Trained ANNs on a working memory task that differentiates monkey and human strategies.
• Analyzed behavioral progression and internal network states during training.

Main Results:

• ANNs demonstrated a behavioral progression from random to recency-like, and finally to selective strategies.
• Apparent recency-like behavior in ANNs emerged as a non-recency-based property of network organization.
• Encouraging recency behavior accelerated learning and enhanced the optimal strategy.

Conclusions:

• Monkeys and humans may represent different stages in the same cognitive learning progression.
• Apparent recency strategies can be emergent properties, not necessarily true recency-based decision-making.
• ANNs offer insights into cognitive discrepancies and can serve as efficient training models.