ABSTRACT
Despite growing concerns about algorithmic bias, most research has focused on improving model accuracy, with less attention given to how human–algorithm collaboration can mitigate biased outcomes. This study adopts a theory-driven approach grounded in the elaboration likelihood model to investigate whether increasing individuals’ motivation to scrutinize algorithmic predictions can reduce bias without sacrificing accuracy. Across two experiments using real-world data, we examined the effects of incentivizing individuals to identify and correct biased algorithmic predictions. Results show that a bias-incentive intervention reduced bias while maintaining predictive accuracy in comparison to a control condition (Study 1). Moreover, the effectiveness of a bias-incentive intervention over an accuracy-incentive intervention was moderated by individual differences in need for cognition (NFC)—a trait reflecting one’s motivation to engage in effortful cognitive processing (Study 2). Participants high in NFC responded more strongly to the intervention, achieving greater reductions in bias.