Prof. Avi Goldfarb - Communicating Uncertainty Can Increase AI Adoption

This study investigates the impact of communicating the uncertainty associated with Artificial Intelligence (AI) recommendations on the adoption of AI by human participants. Using a randomized control trial in an incentive-compatible pricing environment for short-term rental listings, we show that communicating the uncertainty of the outcome of the AI's price recommendation increases the AI's impact on participants' pricing decisions, improving their average revenue by 29\%. In contrast, communicating the uncertainty of AI's model has minimal effects on subjects' pricing decisions and revenues. We estimate a structural model to recover the mechanism behind the effects of communicating uncertainty. We find that subjects often have inaccurate beliefs about the elasticity of demand, and communicating uncertainty about AI's outcome improves their beliefs about elasticity. Our results have managerial implications for improving AI adoption in high-stake decision settings where users' trust in AI is critical for adoption.