# A reinforcement learning and sequential sampling model constrained by gaze data

## TL;DR

This paper introduces a model that combines reinforcement learning and eye gaze data to better predict choices and response times in repeated decision tasks.

## Contribution

The model integrates learned option values and gaze data to enhance predictive accuracy in decision-making.

## Key findings

- The model captures gaze biases on choice and response time.
- It reveals individual differences in absolute versus relative valuation.
- The model successfully explains empirical effects in two eye-tracking experiments.

## Abstract

Reinforcement learning models can be combined with sequential sampling models to fit choice-RT data. The combined models, known as RL-SSMs, explain a wide range of choice-RT patterns in repeated decision tasks. The present study shows how constraining an RL-SSM with eye gaze data can further enhance its predictive ability. Our model allows learned option values and relative gaze to jointly influence the accumulation of evidence prior to choice. We evaluate the model on data from two eye-tracking experiments (total N = 133) and test several variants of the model that assume different mechanisms for integrating values and gaze at the decision stage. Further, we show that it captures a variety of empirical effects, including gaze biases on choice and response time, as well as individual differences in absolute versus relative valuation. The model can be used to understand how learned option values interact with visual attention to influence choice, joining together two major (but mostly separate) modeling traditions.

When people are deciding between options they have encountered in the past, their preferences are largely based on their memory of past experiences with those options. However, simply looking at an option longer can increase the likelihood that it will be chosen, regardless of how rewarding the option is to the decision maker. Most theories of experience-based decision making do not account for this gaze effect. We introduce a tractable and scalable computational model that can predict choices and response times in repeated decision tasks while simultaneously accounting for gaze effects. The model can be used to understand the interplay of learning and visual attention in experience-based decision making.