Joint Human-Automation Decision Making in Road Traffic Management

In this paper we explore automation bias in terms of joint decision making between humans and automation. In an experiment, participants made decisions, and indicated the reason for this decision, in a road traffic monitoring task with the aid of automation of varying reliability (i.e., 25% or 81%). Reliability level had a clear impact on the user’s behavior: at low reliability, participants ignored automation suggestion and rely on their own decision making, whereas in the high reliability condition, participants tended to accept the automation suggestion (even if this was incorrect). Overall, performance is higher as a result of the human intervention that would be expected from automation alone, i.e., accuracy is in the region of 87-96% on all conditions. Performance is affected by the order in which the human and automation give their answers and how much detail they are required to provide. We consider these results in terms of a theory of joint decision making. [156 words]