How working memory and reinforcement learning interact when avoiding punishment and pursuing reward concurrently

Area of Science:

• Cognitive Neuroscience
• Behavioral Psychology
• Computational Psychiatry

Background:

• Human adaptive behavior relies on reinforcement learning (RL) and working memory (WM) systems.
• The interaction between RL and WM during simultaneous reward and punishment learning is poorly understood.
• The impact of psychiatric symptoms on these learning systems requires further investigation.

Purpose of the Study:

• To investigate the interplay between RL and WM in learning to avoid punishment.
• To examine how WM influences RL during concurrent reward and punishment learning.
• To assess the effects of depression, anxiety, and rumination on these learning processes.

Main Methods:

• A novel reward/punishment RL-WM task was administered to an online sample (N=298).
• Computational modeling was used to analyze learning patterns and system interactions.
• Interleaved test phases assessed RL-based retention and WM's influence on RL.

Main Results:

• Working memory (WM) facilitated immediate punishment avoidance, but this was poorly retained by the reinforcement learning (RL) system.
• Evidence of WM blunting RL was observed midway through learning, but this effect diminished after further learning.
• Individual differences modulated the WM-RL interaction, with some individuals showing WM facilitation of retention.
• Task performance remained largely intact despite depression, anxiety, and rumination.

Conclusions:

• The WM system supports short-term punishment avoidance, while the RL system shows limited retention of punishment information.
• Individual differences significantly influence the degree to which WM affects RL.
• Behavioral performance on this task is resilient to internalizing psychiatric symptoms like depression and anxiety.