Intelligent support for Human Oversight: Integrating Reinforcement Learning with Gaze Simulation to Personalize Highlighting
This work addresses the challenge of improving situation awareness for users in time-critical monitoring tasks, though it appears incremental as it builds on existing RL and gaze simulation methods.
The paper tackled the problem of designing interfaces for human oversight by using reinforcement learning to personalize alerting strategies, balancing critical event highlighting with cognitive costs, and found that RL-based highlighting outperformed static rule-based approaches in a delivery-drone oversight scenario.
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.