In this talk, Kalinda Ukanwa from the Marshall School of Business of University of Southern California investigates conditions under which algorithmic bias can impact long-term demand and profits. Her research demonstrates that although there can be a short-term profit advantage from biased algorithms, non-biased algorithms earn higher long-term profits, on average, when consumer word-of-mouth and competition are factored in. This research finds that the long-term benefits of switching to an algorithmic design which ignores group identity information and incorporates word-of-mouth considerations in the algorithm’s objective function. However, for firms that must persist in using group identity information, this research recommends increasing investment in methods of measurement error reduction and increasing exposure to consumers of different populations. By doing so, a firm could reduce algorithmic discrimination in service while improving its long-term profits and societal well-being.
This talk was presented at the Harvard Business School Crossing Disciplines: Studying Fairness, Bias, and Inequality in Management and Decision Sciences Research workshop on May 21, 2021.
For more on the Digital Initiative, check out digital.hbs.edu.
7 июл 2024