Testing models at the neural level reveals how the brain computes subjective value.

Area of Science:

• Neuroscience
• Cognitive Science
• Decision-Making Research

Background:

• Subjective value is central to decision-making but is a theoretical construct.
• Distinct computational models of subjective value often yield similar behavioral predictions, hindering mechanistic understanding.
• Behavioral data alone is insufficient to differentiate between competing decision-making models.

Purpose of the Study:

• To investigate how different computational models of subjective value are implemented in the brain.
• To differentiate between competing decision-making models using neural data.
• To advance mechanistic understanding of subjective value computation.

Main Methods:

• Model comparison at the neural level.
• Analysis of brain regions involved in subjective value computation.
• Distinguishing between models based on reward distributions and probability distortions.

Main Results:

• Distinct theoretical models of subjective value computation show differential neural implementation.
• Frontal cortical regions implement a model based on reward statistics.
• Intraparietal cortex and striatum implement a model based on probability representation distortions.

Conclusions:

• Neural model comparison offers deeper mechanistic insights into decision-making than behavioral analysis alone.
• Different brain regions utilize distinct computational strategies for subjective value.
• This work provides a framework for understanding the neural basis of choice computation.