Real Autonomous Driving from a Passenger’s Perspective: Two Experimental Investigations Using Gaze Behaviour and Trust Ratings in Field and Simulator

Trusting autonomous vehicles is seen as crucial for their dissemination. However, research on autonomous driving so far is restricted by using closed training courses or simulators and by comparing behaviour and evaluation while driving oneself (a manual car) with being driven (by an autonomous car). In the current study, we investigated passengers’ eye movements, categorized as safety-relevant or not safety-relevant, and trust ratings while being driven, once manually and once by an autonomous car, in real traffic as well as in a simulator. As some of the effects observed in the field experiment might have been caused by driving style, driving style was additionally varied in the simulator. Fixations in safety-relevant regions (e.g., on the road and steering wheel) were observed more frequently during safety critical driving situations than during regular driving. More safety-relevant fixations for the autonomous compared to the manual driving mode were observed particularly in the field. Trust ratings were affected by driving mode mainly in the simulator: Here, being driven autonomously led to a lower reported trust than believing to be driven by a human driver. Driving style showed to affect trust ratings, but not gaze behaviour in the simulator experiment. Correlations between gazing into safety relevant regions and trust ratings were of smaller descriptive size than in recent investigations on drivers, suggesting that gazing into safety-relevant regions as objective alternative to trust ratings may not be as exhaustive for passengers as for drivers