Reinforcement Learning Integrates Gaze-Constrained Sequential Sampling
AI-assisted, source-derived brief produced by the Let's Data Science Automated News Desk. The source material used is linked on this page.
- Source event:
- first reported
- LDS brief:
- publication time is not available in the public LDS lifecycle record
On March 6, 2026, Hayes and Touchard publish in PLoS Computational Biology an RL-SSM constrained by eye gaze that jointly models learned option values and relative gaze to predict choices and response times, evaluated on two eye-tracking experiments (N=133). The paper compares additive and multiplicative integration mechanisms, captures gaze-driven choice and RT biases and individual valuation differences, and provides data and code on GitHub.
Key Points
- 1Introduces RL-SSM that integrates learned option values with trial-level gaze to influence evidence accumulation.
- 2Demonstrates that gaze improves predictive accuracy, explaining choice and response-time biases across two experiments (N=133).
- 3Provides a tractable, scalable model and public code enabling practitioners to incorporate attention into RL analyses.
Scoring Rationale
High novelty and practical code availability drive score, with moderate scope limited to decision-modeling and lab-based eyetracking.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems