Learning efficient representations of environmental priors in working memory.

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Psychology

Background:

• Experience shapes expectations and environmental learning through inference models.
• Working memory can exhibit biases, such as in color estimation, influenced by learned priors.
• These biases may stem from synaptic modulation in neural circuits based on stimulus history.

Purpose of the Study:

• To investigate the neural mechanisms underlying systematic biases in human working memory.
• To determine if experience-dependent learning models better explain working memory biases than fixed-bias models.
• To explore how inferred environmental structure influences cognitive strategies.

Main Methods:

• Analysis of human response data from a delayed-estimation task involving color stimuli.
• Comparison of experience-dependent learning models against models with fixed biases.
• Modeling of neural circuits and synaptic connectivity modulation.

Main Results:

• Most subjects' response distributions were better explained by experience-dependent learning models.
• Systematic biases in working memory align with common stimulus values due to neural activity attractors.
• Stimulus distributions in the task did not always match population biases, highlighting individual learning.

Conclusions:

• Systematic limitations in working memory reflect efficient representations of inferred environmental structure.
• Synaptic connectivity modulation based on experience shapes neural representations in working memory.
• This research offers insights into how humans integrate environmental knowledge into cognitive strategies.