Abstract
In this paper, a collision-prevention support alarm at road intersections is used as an example, and a method for analyzing the collision-mitigation ratio is discussed based on the "integrated error of driver and system" that the authors propose. This integrated error refers to not only the system reliability but also to the driver's reliability which will be changed by driver's trust in the system and system dependence during the use of a driving-support system. The investigation regarding the effects of these driver's mental states on collision mitigation considering the integrated error demonstrated that it is important to design an optimal human-machine interface that reduces the driver dependence on the system, while keeping the sense of trust high in order to reduce the number of collisions. However, the study on the trust and dependence found a positive correlation (r= 0.66) between these two factors. This means that it is not easy to satisfy this optimal condition and that developing a method for satisfying this optimal condition in order to reduce the risk-taking behavior of drivers is an important issue.