Description: Department Seminar: Joel Waldfogel (University of Minnesota)

Location: 368A Heady Hall

Contact Person: Donghyuk Kim

Title: The potential welfare benefits of AI in a differentiated product market

Abstract: The benefit that consumers derive from experience goods depends on the information they have prior to purchase. Artificial intelligence (AI) and rich consumer data have delivered new prediction technologies, with the promise of steering consumers toward more suitable products; but the size of the potential benefits from better prediction is unknown. Using cumulative usage time data on 50,000 users and 100 popular games on the video game platform Steam, we quantify AI’s potential benefit by creating personalized predictions of how much time each user would spend with each game. We use these predictions, along with a simple utility maximizing model of consumers, to measure potential welfare effects of better predictions on buyers and sellers. Using both a Cobb Douglas calibration and a logit model of bundle choice, we find large effects: AI-based predictions, if heeded, would raise consumer surplus by roughly two thirds of current expenditure while reducing expenditure by about a fifth.