Autonomous vehicles are complex, technologically opaque, and can vary greatly in what perceptual capabilities they are endowed with. Because of this, it is reasonable to expect people to have difficulties in accurately inferring what an autonomous vehicle has and has not seen, and also how they will act, in a traffic situation. To facilitate effective interaction in traffic, autonomous vehicles should therefore be developed with people’s assumptions in mind, and design efforts should be made to communicate the vehicles' relevant perceptual beliefs. For such efforts to be effective however, they need to be grounded in empirical data of what assumptions people make about autonomous vehicles' perceptual capabilities. Using a novel method, the present study aims to contribute to this by investigating how people's understanding of the visual capabilities of autonomous vehicles compare to their understanding of those of human drivers with respect to (Q1) what the vehicle/driver can and cannot see in various traffic situations, (Q2) how certain they are of Q1, and (Q3) the level of agreement in their judgement of Q1. Additionally, we examine whether (Q4) there is a correlation between individual differences in anthropomorphizing and Q1. The results indicate that people generally believe autonomous vehicles and human drivers have the same perceptual capabilities, and that they therefore are subject to similar limitations. The results also indicate that people are equally certain of their beliefs in both cases, strongly agree in both cases, and that individual differences in anthropomorphizing are not associated with these beliefs. Implications for development of autonomous vehicles and future research are discussed.

Autonomous vehicles are regularly predicted to be on the verge of broad integration into regular traffic. A crucial aspect of successful traffic interactions is one agent’s ability to adequately understand other agents’ capabilities and limitations. Within the current state of the art concerning self-driving cars, there is a discrepancy between what people tend to believe the capabilities of self-driving cars are, and what those capabilities actually are. The aim of this study was to investigate whether people attribute the capacity of object permanence to self-driving cars roughly in the same manner as they would to a human driver. The study was conducted with online participants (N = 105). The results showed that the participants did not attribute object permanence differently between a self-driven car and a human driver. This indicates that people attribute object permanence similarly to self-driving cars as they do toward human drivers. Furthermore, the results indicate no connection between participants’ tendency to anthropomorphize and whether they attributed object permanence or not. The findings provide evidence for the issues connected to the perceptual belief problem in human-robot interaction, where people attribute capabilities to autonomous vehicles that are not there. The results highlight the importance of understanding which mechanisms underlie these attributions as well as when they happen, in order to mitigate unrealistic expectations.