Assessing Human Use of Partially-Automated Motor Vehicles: A Subjective Method for Use by Human Factors Experts

Abstract

This paper describes a pilot test conducted of a quick assessment method in which experts evaluated the safety and usability of partially autonomous vehicles. Two PhD human factors experts drove 3 2018-model-year, partially-automated vehicles (Tesla Model X, Subaru Legacy, and Mazda CX5) and identified safety and usability problems related to the driver interface. From multiple drives of each vehicle over a well-defined course, comments were obtained, and they were classified into 5 safety/usability categories (e.g., not sure what the vehicle did, not sure why it did what it did) with 5 severity levels (cosmetic to catastrophic) in each category. The method was feasible. Planning the route and instrumenting the vehicles proceeded without difficulty, as did the coding of the data. Improvements such as using text-to-speech software (instead of time-consuming manual transcription) to support transcript generation, alternative problem categories, and revised problem severity scoring are described.