Contrast bias, not confirmation bias, shapes daily impression updating across social and non-social targets.

Previous literature on impression updating has paid extensive attention to confirmation bias - the tendency to seek out, favour, interpret or remember information in a way that reinforces existing beliefs. Fewer studies examined the contrast effect/bias in the sequential updating of impressions. Contrast bias occurs when the difference between current beliefs and new information is exaggerated, leading to a stronger shift in evaluation than the new information alone would justify. In this work, we conducted two Ecological Momentary Experiments to capture real-world impression formation and updating on both social (e.g. an employee) and non-social (e.g. a vacation house) targets. We also developed novel computational models of biased impression updating based on parametric distortions of Bayesian ideal observers. Across the two studies, we found that daily impression updating exhibited significant contrast bias, with negative contrast bias exerting a stronger influence than positive ones. Additionally, we found no evidence for differences in impression updating between social and non-social targets. By focusing on within-person impression updating over time, these findings reveal contrast bias as a core dynamic of impression updating, systematically shaping how impressions change in response to inconsistent information.