Are Investment Decisions Suboptimal When AI is the Source of Financial Information?

• In Event: Investor Response to Gender and Technology

Abstract

An increasing number of businesses are using artificial intelligence (AI) software in their accounting systems to reduce uncertainty and improve accuracy. However, algorithm aversion theory (Dietvorst, Simmons, and Massey 2015) indicates that individuals often avoid information provided by automated systems as compared to that provided by humans, suggesting that the use of AI might not result in the anticipated benefits. We investigate how disclosing the use of AI rather than human staff for estimating the fair value of an asset influences investment decisions through decreased levels of attentiveness. Consistent with algorithm aversion, we find that disclosing the use of AI rather than the use of human staff to estimate the asset’s fair value dampens investor responses. Specifically, the use of AI reduces additional investment when the fair value information is positive, as well as reduces investment withdrawal when the fair value information is negative. We also find that feelings of attentiveness mediate the effect of information sources (AI versus human staff) on investment decisions. Our study has practical implications for regulators and managers exploring the effectiveness of more widespread use of algorithms in accounting systems. Our study also has theoretical implications by identifying the relationship between affective responses and algorithm-aversion behavior.

Authors

• Tom Downen, UNC Wilmington

• Sarah Kim, National Cheng Kung University

• Lorraine S Lee, University of North Carolina-Wilmington