The Effect of Perceived Task Role on Reliance on Artificial Intelligence in Unfamiliar Face Matching Tasks

Abstract

Unfamiliar face matching is the process of observing two faces and determining whether they belong to the same person (a match) or two different people (a mismatch). Primarily required in security and identification contexts, this task is surprisingly difficult for humans. With Artificial Intelligence becoming increasingly powerful in automating mundane tasks, current state-of-the-art Automated Facial Recognition Systems (AFRS) can greatly outperform their human counterpart; however, they still often require human supervision and/or input. The 'human-machine interaction' is a term that describes the way humans and machines, in this case AFRS, function together. Whilst the impact of factors such as perceived responsibility and self-reliance on behaviour has been observed with respect to between-human interactions, their effect on the human-machine interaction remains mostly unexplored. This study aims to explore whether manipulating the perceived role in the human-machine interaction can affect trust in automation, complacency, automation-reliance, and ultimately performance in an AFRS-assisted unfamiliar face matching task. Whilst we observed a clear increase in performance when AFRS-assistance was introduced, we found no significant change in performance or trust based on perceived role. Furthermore, human operators curtail the performance of the AFRS regardless of their perceived role in the human-machine interaction. Keywords: Face Matching, Perceived Role, Trust in Automation, Human-Machine Interaction