Australasian Mathematical Psychology Conference 2019

Changing our minds about change-of-mind models: Existing models cannot account for effects of absolute evidence magnitude

William Turner
Melbourne School of Psychological Sciences, University of Melbourne
Daniel Feuerriegel
Melbourne School of Psychological Sciences, University of Melbourne
Milan Andrejevic
Melbourne School of Psychological Sciences, University of Melbourne
Robert Hester
Melbourne School of Psychological Sciences, University of Melbourne
Stefan Bode
Melbourne School of Psychological Sciences, University of Melbourne

To navigate the world, we rely heavily on our ability to make accurate perceptual judgements. However, errors of judgement do inevitably occur. In these situations, rapid ‘changes of mind’ are required to correct or abandon ongoing actions. Recently-developed computational models posit that perceptual judgements are made when evidence is accumulated to a given criterion, and changes of mind occur if evidence later favours a different response. Critically, current models diverge in their predictions regarding the effect of absolute evidence magnitude (i.e., the sum of evidence for opposing choices) on changes of mind. The current study therefore investigated whether absolute evidence magnitude influences the rate and latency of changes of mind. Participants (n=30) indicated which of two flickering greyscale squares was brightest. Following an initial decision, the stimuli remained on screen for a brief period (1s) during which participants could change their response. To investigate the effect of absolute evidence, the total luminance of the two squares was varied whilst the difference in luminance was held constant. Increases in absolute evidence were associated with faster, less accurate initial responses, replicating previous reports. However, high levels of absolute evidence were also associated with slower, less accurate changes of mind (i.e. fewer errors were corrected, but more correct responses were spoilt). The pattern of initial responses can be explained by the presence of input-dependent noise in the decision process, which varies either within or across trials. However, the pattern of change-of-mind responses cannot be fully accounted for by existing models, challenging our current understanding.