How to set the switches on this thing.

Area of Science:

• Computational neuroscience
• Cognitive psychology
• Decision science

Background:

• Reinforcement learning (RL) is a key computational framework for understanding decision-making, particularly when emotions are involved.
• Traditionally, RL models external interactions between an agent and its environment, focusing on stimuli and reinforcement.
• Recent research expands RL to include internal processes, where the organism generates its own reward signals and internal actions.

Purpose of the Study:

• To review recent advancements in reinforcement learning (RL) that incorporate internal cognitive processes.
• To explore the integration of cognitive control mechanisms within RL models of decision-making.
• To highlight the shift towards viewing internal actions, like working memory gating, as integral to RL.

Main Methods:

• Literature review of recent studies in reinforcement learning and cognitive control.
• Analysis of theoretical frameworks that redefine the agent-environment boundary in RL.
• Focus on studies incorporating internal states, self-generated rewards, and cognitive control mechanisms.

Main Results:

• Reinforcement learning models are increasingly incorporating internal states and self-generated rewards.
• Cognitive control is being reconceptualized as a key component within internal RL processes.
• Internal actions, such as working memory gating, are now considered on par with external actions in RL.

Conclusions:

• The scope of reinforcement learning is expanding beyond external environmental interactions.
• Internal cognitive processes, particularly cognitive control, play a crucial role in shaping decision-making within RL.
• Future research should further explore the interplay between internal and external factors in affective decision-making.