No tendency for human operators to agree with automation whose response bias matches their own

Abstract

Evidence suggests that false alarm-prone decision aids can engender stronger disuse than miss-prone aids, even when automation false alarms and misses are matched in perceptual characteristics. The present experiment sought to replicate this effect, and examine whether it reflects a tendency for operators to agree with automation whose response bias matches their own. Participants performed a simulated baggage screening task, alone or with assistance from an automated decision aid prone either to misses or false alarms. A point system encouraged participants themselves to adopt either a conservative, liberal, or neutral response bias. Target-present responses were faster from participants assisted by the miss-prone aid than from participants assisted by the false alarm-prone aid, regardless of the human operators' response bias. Neither response times nor accuracy rates, however, showed evidence of a generalised asymmetry in the effects of automation false alarms and misses.