Artificial Intelligence and Algorithmic Pricing in Practice: Machine Learning Methods and Antitrust Implications

Abstract

Algorithmic pricing has emerged as the digital weapon in a world in which antitrust enforcement risks following statutory guidelines crafted for an analog age. This observation does not imply that existing antitrust laws do not address the role of algorithmic pricing in anticompetitive conduct arising from horizontal, vertical, or hub-and-spoke agreements. Rather, it reflects acknowledgment that technological advancement now affords anticompetitive conduct a greater degree of camouflage that strains detection through traditional antitrust methods. 2 As this essay discusses, algorithmic tacit collusion has the potential to render stalwart antitrust defenses against explicit coordination little more than a Maginot Line in industries that have adopted algorithmic pricing as well as the potential to create an adverse selection problem that drives out potentially beneficial applications. 3. This essay addresses three main applications of such pricing: price discrimination, rivals using different algorithm providers, and rivals using the same algorithmic provider, with examples of use on E-commerce platforms such as Amazon and Walmart, airlines, real estate, and hospitality industries. Concurrently, this essay addresses the role of anchoring on acceptance and rejection of pricing recommendations. Next, in an effort to "demystify" algorithms such as Qlearning, neural networks, random forests and others, this essay offers an accessible introduction to these algorithms and their applications. In each case, this essay discusses the key inputs and likely outputs of interest from these algorithms for legal practitioners and regulators. This essay concludes with an overview of explainable artificial intelligence (XAI) which consists of various methods that attempt to offer insight into black-box pricing algorithms.

Keywords: antitrust, artificial intelligence, machine learning, AI, algorithmic pricing