The uptrend in businesses’ use of artificial intelligence (AI) now includes algorithm-based pricing, which is raising concerns about harm to competition. More and more companies have been implementing algorithm-based pricing mechanisms in their operations. Algorithm-based pricing is not formulated solely by people, but rather involves a machine (computer) that is, at least partially, setting prices. For the most part, a “dominant criterion” governing the pricing process is defined for the algorithm, such as maximizing profit, maximizing revenues, etc. Advanced algorithms of this type are based on artificial intelligence, which enables them to make the decisions that implement the criterion defined for them at every decision junction, while constantly learning.

Companies are very well aware of the fact that these new pricing mechanisms will improve their results and increase their profits. For example, recent media reports note that Israir Airlines installed a dynamic pricing engine that boosted its profits significantly.

When it comes to pricing, algorithms offer significant advantages. AI-based algorithms are capable of performing complex analyses easily, efficiently, and quickly by processing an enormous volume of information. Their computational power far exceeds human capabilities when it comes to collecting, organizing, and analyzing information. This enables the reaching of decisions much faster and more accurately according to changing market conditions, all in real time. Another major advantage is that algorithms are capable of analyzing a company’s price history and, perhaps more importantly, the prices and price history of competitors, as well as their behavior under changing market conditions over time. This thus enables rapid and agile responses.

Pricing Engines and the Free Market

The adoption of such pricing engines may be a blessing, as they enable companies to better utilize their inventories and production capacity. However, at the same time, they may pose a risk of harm to competition, especially in markets with few competitors. One of the main principles underpinning competition, which leads to a reduction in prices, is a company’s uncertainty about its competitors’ future behavior. This uncertainty does not allow a company to “rest on its laurels” and compels it be efficient, to improve the quality and variety of its products and services, and to reduce prices. Extensive use of algorithm-based pricing engines is likely to reduce these uncertainties, as they enable better and more accurate predictions of competitors’ future behavior, inter alia, by analyzing their past strategies. When the algorithms are programmed to react, to the extent possible, by raising prices, a concern naturally arises that the advantages of these pricing engines will be the downfall of competition. This will ultimately be reflected in higher prices to consumers.

These concerns are likely to materialize rapidly in concentrated markets with few players, where competition failures already exist. In concentrated markets, companies can already predict the behavior of their competitors relatively easily, and thus behave in a less competitive manner and charge higher prices than would be charged under competitive conditions, even without explicit coordination between competitors. Now, algorithm-based pricing engines will make it even easier for players in concentrated markets to predict the behavior of their competitors. The ability to analyze competitors’ behavior and identify the characteristics of competitors’ pricing engines and their modes of operating will enable such companies to raise prices and maximize profits.

Unlawful Coordination between Pricing Algorithms

Technological progress cannot and should not be stopped. Therefore, the question arises of whether, and under what circumstances, competition law can intervene to prevent pricing algorithms from harming competition. And, if so, what tools are available for this purpose?

Within this context, it is important to focus on the distinction in competition law between “explicit coordination” (or “explicit collusion”), concerted practices and “parallelism” (also known as “tacit collusion”). In general, explicit coordination occurs when competitors reach an understanding or agreement about competitive issues between them (prices, quantities, customers, operating regions, etc.). In the vast majority of instances, such explicit coordination is prohibited, and may be considered a crime in certain jurisdictions.

In the world of algorithm-based pricing engines, it is possible to raise an allegation of explicit coordination (or a less explicit coordination known as “concerted practice”), constituting a prohibited restrictive arrangement that harms competition and consumers. One such scenario is if competitors program their algorithms to “communicate” with each other and exchange information about the algorithms they are running, especially about the algorithms’ properties and the dominant criterion governing how they operate. (Theoretically, one algorithm can be programmed to raise prices in response to a competitor hiking prices, while another algorithm can be programmed to lower prices in response, in order to grab market share.) Another scenario is competitors agreeing to purchase the same algorithm with the same features. In such instances, competition authorities can employ the “classic” tools for dealing with cartels, including the filing of indictments (where applicable).

Algorithms Learn Market Conditions

Another possible challenge is a scenario whereby the AI-based algorithm constantly “learns” market conditions and optimizes its operations to achieve the goals defined for it by independently learning to communicate directly with algorithms operated by competitors, even though it has not been programmed to do so. Could such “machine” communication constitute a prohibited restrictive arrangement? If so, can a corporation and its officers be held liable for the coordination created by the AI-based algorithm or for not preventing it at the outset? Furthermore, can the developer and programmer of the algorithm be held liable in such instance? These are extremely complex questions in uncharted territories that address the clash between the public’s interest in competition and the limits of protection of individual rights, including the extent to which a company or officer can be held criminally liable for acts committed by an AI-based tool used by that company.

The situation becomes even more complex when it comes to “adjustments” or “parallelism”, i.e., when competitors are not coordinating competitive issues between them, but are merely adjusting their behavior, independently, according to their competitors’ behavior. Such adjustments (which are not prohibited under competition law) may harm competition and the public, especially in oligopolistic markets with few competitors. As noted above, the use of algorithm-based pricing mechanisms may increase the risk of such harm.

Competition Laws Need to Delegate Appropriate Powers

Unless competition laws are amended to delegate appropriate powers, competition authorities will continue to have a hard time preventing competitive concerns in such instances. In Israel, the Israel Competition Authority’s Director General has been vested with an authority almost unique in the world. Namely, the authority to declare competitors in oligopolistic markets (or in markets with few rules of competition) a “concentration group,” and to take measures against the members of the group that may prevent harm or concerns of significant harm to the public or to competition, or to take measures to significantly increase competition in a specific sector. The director general may order, inter alia, a member of the concentration group to discontinue a particular activity if that activity facilitates coordination among the other members of the group.

Theoretically, the director general may exercise this authority in instances of growing concerns of harm to competition resulting from use of algorithm-based pricing in markets with few competitors (or in markets characterized by competition failures). It is important to note that, although the director general was delegated this authority more than a decade ago, it has hardly been exercised to date. So far, only one concentration group has been declared, and its members were issued orders. It will be interesting to see if this technological development in pricing, which may be used by companies to legally raise prices using algorithms, will trigger higher use of this tool.

In any case, there is no doubt the use of algorithm-based pricing engines poses new and complex challenges for competition authorities. In due time, which may not be as far off as one might think, competition authorities are likely to be called upon to make difficult legal decisions in the scenarios described above or in similar ones.