Confirmation bias in visual search
53.315, Tuesday, 20-May, 8:30 am - 12:30 pm, Jacaranda Hall
Jason Rajsic1, Daryl Wilson2, Jay Pratt1; 1Department of Psychology, University of Toronto, 2Department of Psychology, Queen's University
In this study we tested whether or not confirmation bias, a well-known decision-making bias consisting of a tendency to selectively search for and evaluate information expected to confirm a focal hypothesis, can occur in visual search. Participants completed visual searches for a target letter, and were asked to make one response when the target letter appeared in a specified color and another response when the target letter appeared in a different color. The set size was held constant at eight, and the critical manipulation was the proportion of the stimuli that were in the specified color, referred to as the proportion of “hypothesis confirming” (HC) stimuli, compared to the unspecified color, referred to as the “hypothesis disconfirming” (HD) stimuli in a given search display. In Experiment 1, we found a confirmation bias, as participants searched through HC stimuli first, even when that required searching more items than searching through the smaller HD set. In Experiment 2, we attempted to attenuate the confirmation bias by incorporating a color preview display prior to the visual search display, so that participants could plan their search in advance of the stimuli appearing, therefore allowing them to implement an unbiased strategy. The results showed that the bias was attenuated, although participants were not able to detect the presence of a target letter in the HD set as efficiently as determining the absence of a target letter in that same set. Overall, these findings suggest that visual search is susceptible to confirmation bias and that this bias can be diminished by cognitive control mechanisms. Furthermore, this work shows that visual search can be used as a model for determining the role that attentional mechanisms may have in generating and maintaining confirmation biases.