Accounting for multiscale processing in adaptive real-world decision-making via the hippocampus

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Neuroscience

Background:

• Value-based decision-making relies on reinforcement learning models, primarily model-free (MF) and model-based (MB).
• Existing research emphasizes striatal and prefrontal regions for
• Purpose_of_the_Study
• Main_Methods
• Main_Results
• Conclusions

Purpose of the Study:

• To investigate the hippocampus' role in value-based decision-making.
• To propose the successor representation (SR) as a mechanism for hippocampal state encoding.

Main Methods:

• Review of neuroscientific and computational literature.
• Analysis of research linking hippocampal sequences to SR models.

Main Results:

• Hippocampal sequences, when implemented in reinforcement learning agents, enhance performance.
• SR models enable agents to perform biologically plausible multiscale temporal processing.

Conclusions:

• The hippocampus plays a crucial role in decision-making beyond memory and spatial learning.
• The successor representation offers a framework for understanding multiscale temporal processing in decision-making.
• Integrating hippocampal mechanisms advances current models of value-based decision-making.