Acquiescence Bias in Aided Visual Search

Abstract

This study looked for evidence of acquiescence bias, a tendency to agree rather than disagree, in human-automation interaction. This bias has not previously been identified in automation use and its existence could help explain operator mistakes made when interacting with automated decision aids. In the first two experiments, it was as beneficial to agree with the aid in one condition and it was beneficial to disagree with the aid in the other. The third experiment gave participants a 50% accurate aid, which offered them no meaningful help. Experiments 1A and 1B revealed that participants performed better when it was beneficial to agree rather than beneficial to disagree with the aid. Experiment 2 revealed that the aid influenced participants’ search strategies even though they were told that the aid was as helpful as a coin flip. The findings suggest that participants display acquiescence bias when making decisions with an automated aid.