Decoding Driving Words: A Holistic Exploration of Older Drivers’ Early-stage ADAS Adaptation Using Structural Topic Modeling on Naturalistic Driving Data

Abstract

This study examined how older drivers adapt to Advanced Driver Assistance Systems (ADAS), focusing on trust, satisfaction, and behavioral patterns during early use. Using Symbolic Aggregate Approximation and Structural Topic Modeling (STM), we analyzed naturalistic driving data from participants aged 70 to 79 over 5 weeks. STM identified driving behavior topics, and their prevalence was modeled against weeks of ADAS exposure and weekly trust/satisfaction ratings for Adaptive Cruise Control (ACC) and Lane Keeping Assist (LKA). The results revealed that (1) increasing prevalence of active steering inputs on highways was associated with continued ADAS exposure, (2) higher ACC trust and satisfaction were associated with decreasing prevalence of active speed adjustments on highways and (3) increasing steering in mixed-speed environments, (4) decreasing prevalence of active speed adjustments on highways was associated with higher LKA trust. These findings can inform policy and training to support appropriate ADAS use among older drivers.