AI algorithms are increasingly being used in markets. It is generally agreed that they create a wealth of value by facilitating economic interactions. However, there are also concerns that when applied to market environments, they could lessen competition and harm consumers.
 This proposal sets forth an important first step of an articulated research program on AI and competition. Methodologically, these projects rigorously combine ingredients from economics and computer science (CS), departing from the stylized modelling approach of the literature and instead are aimed at analysing state-of- the-art algorithms developed in the CS literature.

 The proposal is structured as three organic projects focusing on two important classes of AI algorithms. The first class, price-setting algorithms, are tools that are used by online merchants to set retail prices. As the algorithms of each merchant interact in a common market, they can mutually learn to defeat competition. We plan to study how these collusive conducts are learned to propose a direction for potential regulation of these algorithms. The second class, recommender systems, are ubiquitous online algorithms that suggest products and content to consumers that are based on their past behaviour. For example, Amazon.com uses recommender systems to rank products on search result pages, or to suggest other products that the user may be interested in. The aim of the proposed research project is to assess whether RS are pro-competitive or anti-competitive by studying how firms endogenously react to RS in terms of price, product design, and entry/exit decisions. Algorithms and their impact on market outcomes naturally require an interdisciplinary approach. Dealing with actual algorithms requires new tools to supplement the economist’s traditional analytical toolbox. For this reason, an essential part of this research proposal is the creation of a joint research agenda via a “research pod” with an associated scientific program.