Intelligent support for Human Oversight: Integrating Reinforcement Learning with Gaze Simulation to Personalize Highlighting

TL;DR

This paper proposes a reinforcement learning approach combined with gaze simulation to personalize alerting strategies in human oversight interfaces, aiming to improve situation awareness efficiently under time constraints.

Contribution

It introduces a novel integration of gaze behavior models with reinforcement learning for adaptive UI highlighting in oversight tasks.

Findings

RL-based highlighting outperforms static rule-based methods

Gaze simulation effectively enables learning without real-world deployment

Initial results indicate improved attentional support in drone oversight

Abstract

Interfaces for human oversight must effectively support users' situation awareness under time-critical conditions. We explore reinforcement learning (RL)-based UI adaptation to personalize alerting strategies that balance the benefits of highlighting critical events against the cognitive costs of interruptions. To enable learning without real-world deployment, we integrate models of users' gaze behavior to simulate attentional dynamics during monitoring. Using a delivery-drone oversight scenario, we present initial results suggesting that RL-based highlighting can outperform static, rule-based approaches and discuss challenges of intelligent oversight support.