The Effects of Making Independent Judgements and Time Pressure on Human Use of Automated Facial Recognition Systems

Abstract

Automated Facial Recognition Systems (AFRS) are commonly used in airports to help verify the identity of individuals. The use of AFRS in airports is often teamed with human observers who assist in validating inconclusive AFRS decisions. This verification usually occurs under considerable time pressures. Although previous research has explored time pressure effects on human accuracy in face matching tasks, to date, there has been no research conducted on how time pressure influences human reliance on AFRS. Therefore, this research aimed to investigate how humans use AFRS, and whether time pressure affects this relationship. We first validated a new trial procedure which measured reliance on AFRS through tracking identification decision change. For each pair of stimuli, participants first made an independent identification decision. An AFRS decision was then presented on screen, followed by participants submitting a final response. We applied our new trial procedure to a time pressured task. Each participant (n = 56) completed the face matching task in trial blocks where stimuli were presented for 2 seconds, 5 seconds, and 10 seconds. Participant accuracy significantly improved when guided by AFRS compared to independent human decisions. Time pressure produced a non-significant effect on human performance, although, we observed significantly lower mismatch trial accuracy in higher time pressures than in lower time pressures. Furthermore, across both experiments, participants often relied on AFRS, even when the AFRS displayed an incorrect identification decision. Our results have implications for the role of humans performing oversight of AFRS, and potentially limiting the performance of automation. Keywords: face matching, AFRS assistance, time pressure, reliance on AFRS.