Computational Mechanisms Underlying Motivation to Earn Symbolic Reinforcers

Area of Science:

• Neuroscience
• Cognitive Science
• Behavioral Economics

Background:

• Reinforcement learning (RL) explains how agents learn through rewards and punishments.
• Symbolic reinforcers (e.g., tokens) are common but their motivational mechanisms are unclear.
• Understanding how symbolic rewards acquire value is crucial for various applications.

Purpose of the Study:

• To investigate how monkeys learn to maximize fluid rewards using tokens as symbolic reinforcers.
• To model how the value of task states, influenced by multiple features, drives motivation.
• To correlate computational state values with behavioral measures in a token-based task.

Main Methods:

• Developed a Markov decision process (MDP) model to compute state values based on task features.
• Collected behavioral data from 5 monkeys performing a token-based reward task.
• Analyzed fixation times, choice reaction times, and abort frequency in relation to computed state values.

Main Results:

• Behavioral metrics (fixation time, reaction time, abort frequency) significantly correlated with model-derived state values.
• The MDP model successfully captured learning and behavior related to symbolic reinforcement.
• The model provides a framework for predicting neural activity changes based on state value.

Conclusions:

• Symbolic reinforcement, like tokens, can effectively motivate behavior by acquiring value through association with primary rewards.
• Computational modeling of state value is a powerful tool for understanding motivation and decision-making.
• This study offers insights into the neural and behavioral mechanisms underlying symbolic reward learning.