Adaptation to Misinformation: A Drift Diffusion Model Analysis of Prior Likelihood

Abstract

Perceptual decision-making involves rapidly classifying sensory information to select appropriate responses. These decisions can be influenced by choice biases, which are systematic preferences for certain options and may rise from prior expectations of the likelihood of the occurrence of an event. While these biases increase decision efficiency, they can lead to suboptimal outcomes if expectations misalign with reality. In this study, we first investigated whether expected likelihoods of events induce proportional choice biases. Then, we examined the extent to which discrepancies between expected and actual likelihoods influence the development of these choice biases. We were particularly interested in any potential differences between adaptation to prior information inaccurately predicting a high likelihood of an event versus inaccurately predicting a low likelihood of an event. Participants (N=61) performed a visual discrimination task where prior information predicted the likelihood of occurrence for two options. We manipulated both the strength of the probability and prediction accuracy of the prior information. We then fitted a drift diffusion model (DDM) to the behavioral data of each participant. The starting point parameters of the DDM indicate that likelihood expectations do not result in proportional biases. Furthermore, participants showed no significant adaptation in response to inaccurate prior information, suggesting that bias development is mainly driven by prior information, regardless of the accuracy of this information.