Nutrient-Sensitive Reinforcement Learning in Monkeys.

Area of Science:

• Neuroscience
• Behavioral Economics
• Computational Biology

Background:

• Reinforcement learning (RL) models explain how animals learn from rewards.
• Canonical RL models lack biological and ecological validity regarding intrinsic reward components like nutrients.
• Animals must adapt foraging choices to acquire essential nutrients for survival.

Purpose of the Study:

• To investigate how monkeys adapt choices for preferred nutrient rewards under varying probabilities.
• To advance the biological and ecological validity of RL models by incorporating nutrient composition.
• To understand the influence of nutrient preferences on learning and decision-making.

Main Methods:

• Monkeys performed a task involving choices with varying reward probabilities and nutrient compositions (sugar, fat).
• A nutrient-sensitive RL model was developed to capture value updating based on individual nutrient components.
• Behavioral data on choices and reward history were analyzed to identify nutrient-specific learning effects.

Main Results:

• Nutrient composition significantly influenced learning and choice behavior in monkeys.
• Monkeys showed nutrient-specific reward history effects, with preferences for sugar and fat strengthening the impact of recent rewards.
• Preferred nutrients were chosen even with lower reward probabilities, indicating nutrient-driven decision-making.
• The nutrient-sensitive RL model successfully explained observed choice behavior by updating nutrient-specific values.

Conclusions:

• Nutrient components are critical factors influencing learning and choice by modulating subjective reward values.
• Extending RL models with nutrient-value functions enhances biological validity and uncovers nutrient-specific learning variables.
• This research highlights the importance of incorporating biologically relevant intrinsic rewards into RL frameworks for greater ecological relevance.