How was my performance? Exploring the role of anchoring bias in AI-assisted decision making

Abstract

Organizations leverage artificial intelligence (AI) to analyze data and support decision making. However, the integration of AI into organizational workflows may introduce unintended biases. Despite the proliferation of AI in organizations, no study to date has juxtaposed the impact of human and AI recommendations on decision making. Using two controlled experiments of 775 managers, we explore the impact of AI and cognitive bias on performance appraisal ratings. In particular, we examine anchoring and adjustment bias and present an effective strategy for mitigating this bias. The findings show managers’ performance ratings are impacted by the presence of an AI recommendation. The source of the recommendation (human or AI) interacted with the anchor (high or low) to influence managers’ rating. In particular, a high-anchor produced different performance ratings for each source. However, when exposed to a low-anchor, supervisors did not produce varied estimates from AI and non-AI recommendations. These findings suggest managers should be aware of the differential effects of anchoring and adjustment bias on organizational decisions. An employee’s performance may be rated differently, not because of the employee’s behavior, but because of the source of the recommendation and the magnitude of the anchor. This paper makes several significant contributions: (1) it is among the first studies to empirically test the presence and salience of anchoring bias in AI-assisted decision making; (2) it presents the consider-the-opposite strategy as an approach to effectively debias the anchoring effects of AI recommendations.