Alex Miller, Operations, Information and Decisions, The Wharton School
Abstract: In recent years, firms have started to incorporate new techniques from artificial intelligence into their decision making processes. However, these techniques come with a novel set of reputational and legal risks relating to algorithmic bias and digital discrimination. As more decisions are handed over to black-box algorithms processing large amounts of data, the easier it is for firms to implement discriminatory practices without even realizing it. As managers and executives evaluate the value of investing in artificial intelligence, it is critical they understand both the risk of these methods and ways they might be able to avoid their potentially undesirable downsides. In this project, we study the potential for algorithmic bias in the context of targeted promotional pricing practices among online merchants. By using data from a set of real-world online experiments, we plan to carefully study the empirical incidence of digital discrimination in this domain, identify key factors which increase the risk of algorithmic bias, and evaluate techniques for avoiding its undesirable consequences. Once completed, our research will be able to provide valuable insight into how firms can get value from AI-enabled targeting systems while minimizing their negative discriminatory effects.