Problem: trust and adoption of avs are limited by ONE-SIZE-FITS-ALL APPROACHes to HUman-AI COMMUNICATION.

Solution: We Aim to improve human-autonomous vehicle interaction through personalized, context-aware, and goal-based communication INSTEAD OF taking a one-size-fits-all approach.

FOR AUTONOMOUS VEHICLES (“SELF-DRIVING CARS”) TO ENGENDER APPROPRIATE LEVELS OF TRUST, THEY MUST BE RESPONSIVE TO THE PRAGMATIC, INFORMATIONAL, AND EMOTIONAL NEEDS OF THEIR PASSENGERS.

Study 1: Predicting Trust From Personal Traits Using Machine Learning

Can we use machine learning to predict a person’s trust, concerns with vehicles, and adoption attitudes based on their traits and experiences?

In this study, we used survey to develop a comprehensive profile for a person, then used machine learning and explainability techniques (SHAP value regression) to derive insights on most important factors predicting a person’s trust profile.

Our models were highly accurate (90% or better) and show several traits and specific concerns with AV that can be used by AV designers to tailor interventions to build trust.

In a pre-post mixed-methods experiment (n = 41), we tested the impact of an AI Coach’s explanatory communications modeled after performance driving expert instructions. We compare how different explanatory techniques impact driving performance, cognitive load, confidence, expertise, and trust via observational learning. Through interview, we delineate participant learning processes.

Results show AI coaching can effectively teach performance driving skills to novices. Efficient, modality-appropriate explanations should be opted for when designing effective HMI communications that can instruct without overwhelming. Differences in how successfully participants learned are attributed to how information directs attention, mitigates uncertainty, and influences overload experienced by participants.

Study 3: Autonomous Vehicle Errors, Context, and Personalization

What happens when autonomous vehicle communications mess up?

We developed a custom, state-of-the-art driving simulator and used it to test how AV explanation errors, driving context characteristics, and personal traits affect a person’s comfort relying on the AV, preference for control, confidence in the AV's ability, and explanation satisfaction (n = 232).

Results emphasize the need for accurate, contextually adaptive, and personalized AV explanations to foster trust, reliance, satisfaction, and confidence. We conclude with design, research, and deployment recommendations for trustworthy AV systems.