AI and Interdependent Pricing: Combination Without Conspiracy?
Joshua P. Davis - University of San Francisco School of Law;
Anupama K. Reddy - Joseph Saveri Law Firm
-- Artificial Intelligence (AI) is already used pervasively—particularly in the marketplace—although the general population may be unsuspecting of much of what it does. AI has made extraordinary strides in recent years and they may soon accelerate dramatically. These technological developments offer many benefits. AI, for instance, can detect cancer more reliably from medical scans than doctors. It may also soon be able to improve the speed and accuracy at which sellers assess market demand and adjust their production to improve efficiency. That could decrease waste, resulting in lower prices and greater output.
But there is also a potential dark side to AI. It may empower market actors to harm competition in ways that were not feasible in the past. For example, it could enable competitive sellers to avoid competition in favor of coordination, allowing them to mimic the supracompetitive prices and infracompetitive outputs of monopolists. AI’s potential thus threatens a core strategy of competition (or antitrust) law: to force market actors to compete with one another even though they would prefer not to do so.
Traditionally, competition law has relied on practical difficulties to prevent market actors from collaborating in the absence of a conspiracy. As a result, the laws of many jurisdictions—including arguably the U.S.—have prohibited price-fixing agreements among competitors but have permitted market actors to engage in parallel pricing. Each market actor could independently attempt to approximate monopoly pricing and output with the hope that its competitors would respond in kind—and perhaps even with an awareness that other market actors are doing the same. This pattern is sometimes called conscious parallelism or interdependent pricing. In permitting this type of conduct, we rely on natural impediments to coordination—on a collective action problem—to limit the adverse effects of interdependent pricing.
Enter AI. It can facilitate strategic behavior that to date has not been possible. Often that is beneficial. When it comes to antitrust, it may not be. AI could be used across a vast array of industries to overcome the challenge of coordination. In antitrust, coordination challenges can be a feature, not a bug. Overcoming them could lead to prices that are pervasively inflated above competitive levels, output that is deflated below competitive levels, large market actors that reap extraordinary profits, and ordinary people who suffer as a result.
A potential solution is to prohibit use of AI to engage in interdependent pricing. Commentators have long speculated that it might be best to ban interdependent pricing anyway. After all, it can cause the same kinds of harms to competition as can horizontal price-fixing conspiracies. Judge Richard Posner suggested such a ban over fifty years ago (Richard A. Posner, Oligopoly and the Antitrust Laws: A Suggested Approach, 21 Stan. L. Rev. 1562 (1969)), although he has since recanted that position. Posner recently explained his current thinking on the issue in reviewing a book by Professor Louis Kaplow, Competition Policy and Price Fixing (Richard A. Posner, Review of Kaplow, Competition Policy and Price Fixing, 79 Antitrust L. J. 761 (2014)). Kaplow offers powerful reasons for his tentative conclusion that current law can and should prohibit interdependent pricing. All of this even in the absence of AI!
In our recent article—AI and Interdependent Pricing: Combination Without Conspiracy?—Anu Reddy and I argue that AI will likely tip the balance in favor of Posner’s original position and Kaplow’s current one. If we want to preserve the benefits of competition—the foundation of competition law—we may well need to ban use of AI by market actors to fix a problem that should be left unfixed. Indeed, as we suggest, that may have been the right approach all along.