Anchoring effects in view transitions of data Visualization:Evidence from an ERP study

Abstract

Recently, as people have to deal with increasingly complex information, the anchoring effect tends to be one issue of concern. It is a phenomenon that prior information might bias cognition and result in estimation deviation. To understand the behavioral characteristics and underlying neural mechanisms of the anchoring effect in visualization, a 2 × 2 within-subject experiment was performed under four conditions: lower anchor with a number, higher anchor with a number, lower anchor without a number and higher anchor without a number. We conducted a pie-chart reading task, which was to estimate the value of the target pie chart. Participants’ perception and judgement under different conditions was examined by using event-related potentials (ERPs) and behavioral metrics. The behavioral results showed that a significant anchoring effect was confirmed only under the anchor with a number condition, indicating the anchor without a concrete number is helpful in mitigating the anchoring effect. The ERPs results indicated that the anchor without a number elicited larger P2 and P300 amplitudes than that with a number. This study contributes to the understanding of the neural processing of view transitions in data visualization by providing electrophysiological evidence of this cognitive process. Moreover, the neuro-design method suggested in this study use P2 and P300 as potential indicators of anchoring bias, which could help information visualization researchers to gain deeper insight on the underlying cause of human performance. The recommendations from this study for the design of information visualization with multiple views might be that avoiding the use of concrete numerical data and providing graphic representations could mitigate the anchoring effect.