Toward overtrustfree advanced driver assistance s

Authors: Makoto Itoh

Publish Date: 2011/10/01

Volume: 14, Issue: 1, Pages: 51-60

PDF Link

Abstract

Avoiding human overtrust in machines is a vital issue to establish a socially acceptable advanced driver assistance system ADAS However research has not clarified the effective way of designing an ADAS to prevent driver overtrust in the system It is necessary to develop a theoretical framework that is useful to understand how a human trust becomes excessive This paper proposes a trust model by which overtrust can be clearly defined It is shown that at least three types of overtrust are distinguished on the basis of the model As an example this paper discusses human overtrust in an adaptive cruise control ACC system By conducting an experiment on a mediumfidelity driving simulator we observed two types of overtrust among the three The first one is that some drivers relied on the ACC system beyond its limit of decelerating capability The second one is that a driver relied on the ACC systems by expecting that it could decelerate against a stopped vehicle It is estimated through data analysis how those kinds of overtrust emerged Furthermore the possible ways for prevention of human overtrust in ADAS are discussed

Keywords: