Dark control: The default mode network as a reinforcement learning agent.

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Neuroscience

Background:

• The default mode network (DMN) is associated with baseline human mental activity and conscious awareness.
• Its high energy consumption suggests a significant, yet not fully understood, overarching function.
• Existing theories propose an evolutionarily adaptive role in future anticipation and experience envisioning.

Purpose of the Study:

• To propose a process model explaining how the DMN implements continuous environmental evaluation and prediction.
• To elucidate the DMN's role in guiding behavior through adaptive action policies.
• To integrate diverse interpretations of DMN function within a unified computational framework.

Main Methods:

• Development of a process model for DMN function.
• Formalization of DMN activity using Markov decision processes (MDPs).
• Analysis of how MDPs optimize action policies via value estimates and vicarious trial and error.

Main Results:

• The proposed MDP framework provides a unified perspective on DMN function.
• This model accommodates previous interpretations including predictive coding, semantic associations, and sentinel roles.
• The model offers parsimonious explanations for recent experimental findings in both humans and animals.

Conclusions:

• The DMN's primary function may be the neural optimization of complex behavior through continuous environmental prediction and evaluation.
• Markov decision processes offer a powerful lens for understanding DMN's adaptive role.
• This process model unifies existing theories and explains empirical observations of DMN activity.