A multi-method approach to understanding drivers' experiences and behavior under partial vehicle automation
archive: archived pipeline: cataloged verified
Summary
AAA Foundation for Traffic Safety technical report (Strayer, Cooper, Sanbonmatsu, & McDonnell, University of Utah). Hybrid longitudinal study of driver behavior and experience under SAE Level 2 partial vehicle automation, integrating three methods: (1) on-road experimental sessions with EEG and Detection Response Task (DRT) before and after a 6-8 week familiarization period; (2) naturalistic driving observations during the familiarization period across daily commutes; (3) periodic surveys of trust, perceptions, and intentions. Five Level 2 vehicles were used (Tesla Model 3, Tesla Model S, Cadillac CT6 Super Cruise, Volvo XC90 Pilot Assist, Nissan Rogue ProPILOT) on two Salt Lake City interstates (I-15 high-volume straight; I-80 curvy mountain). Naive drivers (no prior Level 2 experience) were assessed for changes in workload, visual engagement, automation usage, system warnings, fatigue, fidgeting, secondary task engagement, and attitudes.
Key finding
Drivers paid more attention to the driving environment under Level 2 partial automation than during manual driving in the initial session, but after 6-8 weeks of familiarization showed a significant decrease in attention under automation in the simpler highway environment; spectral EEG (frontal theta, parietal alpha) did not show evidence of decreased workload or engagement under automation, highlighting the importance of multiple measures and varied roadway conditions. Naturalistic data showed automation use >70% of the time, increasing system warnings with experience (more relaxed monitoring strategy), reduced automation use under higher driving demands, no automation effect on fatigue or fidgeting, and growing secondary task engagement over time. Surveys showed automation improved the driving experience, reduced stress, and increased intentions to use and purchase automated vehicles, while drivers remained cognizant of risks.
Methodology
Multi-method longitudinal hybrid design. Part 1 (experimental, on-road): within-subjects 2 (Level: 0 manual vs 2 partial automation) x 2 (Highway: I-15 vs I-80) x 2 (Session: pre- vs post-familiarization) design with EEG (frontal theta, parietal alpha; BIOPAC, Fz/Cz/Pz) and vibrotactile DRT (ISO 17488, RTs 100-2500 ms, hit rate). Part 2 (naturalistic): 6-8 weeks of daily commute driving with in-vehicle video recording and system telemetry; behaviors hand-coded for warnings, fidgeting, fatigue indicators, and secondary task engagement; manual-driving control benchmarks compared with automation segments. Part 3 (survey): bi-weekly questionnaires plus pre/post measures of trust, perceptions, attention, and intentions to use/purchase. Five commercially available Level 2 vehicles used to support generalizability.
Sample size: N=30 (12 female, 18 male; ages 18-55, M=35.7, SD=9.3); all Level 2 naive at enrollment
Quality score: 5 / 5