Exploring the Sensitivity of Eye-Tracking Measures for Benchmarking the Visual Demand of In-Vehicle Information System in the Intelligent Cockpit

Abstract

As the integration of intelligent cockpits in production vehicles grows, understanding the visual demand introduced by In-Vehicle Information System (IVIS) interactions and their potential to cause distractions becomes increasingly crucial. Despite extensive research on IVIS visual demand, limited attention has been given to assessing the sensitivity of eye-tracking metrics employed in measuring this demand. The present study investigated the sensitivity of sixteen aggregated eye-tracking measures involving saccade, fixation, and glance. A total of 171 participants interacted with 19 production IVISs via touch screen to accomplish 17 secondary tasks, during the above experiment their eye movement behaviors were recorded. For sensitivity, we employed the Marginal R-squared value from linear mixed effects models and the feature importance from LightGBM models. The percentage of eyes off road time (PEORT) was found to be the most sensitive metric. In conjunction with correlation analyses, we categorized the filtered measures into five groups and selected the most sensitive metrics from each group to formulate an approach for IVIS visual demand assessment. This study fills a gap in existing research concerning the sensitivity of eye-tracking measures for assessing visual demand. Furthermore, we endeavor to establish a comprehensive guideline for visual demand measurement that aligns with production IVIS systems, ensuring the practical applicability and relevance of our findings in real-world driving scenarios.