Regulators Consider a New Theory on Algorithmic Coordination

Transcript

The following transcript of this discussion was edited for clarity.

The digital revolution gave companies powerful new ways to understand their customers — and over time, many turned to third-party platforms for analytics and pricing. But as dominant providers emerged, competitors increasingly relied on the same tools, feeding data into the same algorithms.

Now antitrust authorities are asking an uncomfortable question: Could that convergence amount to coordination, even if the companies never talked to each other?

I’m Bob Mertz, and I’m here with Goodwin partner Arman Oruc, one of the authors of a recent article on this topic in our Forces of Law series.

Bob Mertz: Arman, welcome.

Arman Oruc: Thanks for having me.

Let’s start with the basic theory. What are regulators actually claiming in these situations?

Companies have always collected data to understand market dynamics, analyze demand, and forecast supply — and they analyze data to adjust prices and manage revenue and costs.

Of course, the methods companies used historically were relatively primitive compared to today. Now companies often use third-party platforms and algorithms to gather and analyze data. And these platforms often give advice on how to price products.

The government — and private plaintiffs — has said that when competitors use the same algorithm, they are, in essence, delegating competitive decision-making to that algorithm. And that creates what looks like a collusive price-setting arrangement. That’s the theory.

It looks collusive because they are relying on the same systems for advice?

The government is saying: Now you have competitors using the same algorithm — and very likely the same dataset because the algorithm is getting information from the same set of competitors. Therefore, there is a concentration, a centralization, if you will, of data. That looks like collusive price setting.

Again, that is a theory.

The counterarguments are plenty, and some of them are self-evident. There is no immediate collusion here. There is no doubt about that in most of the cases we have seen. The fact that there might be an algorithm that helps you do the math more precisely doesn’t immediately raise antitrust concerns.

Does that have to do with the fact that there’s no intent to collude?

Right, there is no intent to collude. In none of these cases has there been any allegation that defendant one picked up the phone and called defendant two. There is no allegation that a defendant, as a condition of using an algorithm, told the algorithm-maker that it will only use the algorithm if the maker also gets its competitors to use it. We’ve seen situations like that in the past in other contexts. There’s plenty of precedent in which the conspiracy issues have been explored by the courts. That’s not the case in these situations.

The problem is most acute when the product, the algorithm, is really good. If there’s a good product, people want to use it. It’s a problem for antitrust enforcement if an innocent decision to use a product that is really helpful for your business gets you into antitrust trouble. These are very innocent decisions about trying to do the right thing for your business, your shareholders, and your employees — and trying to run an efficient business.

Where do you see this going? Seattle and San Francisco have already banned algorithmic pricing tools in housing. Do you expect to see more of that in other cities and in other contexts?

In more politically charged industries, we will probably see more legislative action aimed at algorithms. Housing jumps out as the most obvious area. New York just passed a similar law. It’s being challenged.

But there’s no stopping this. These algorithms are extremely efficient. If history and economics teach us anything, it’s that efficiency typically wins.

To ban these algorithms outright is probably bad policy. It won’t work, and that’s a good thing for everybody. You do want efficient markets. Efficiency requires good decision-making by all competitors. I’m not suggesting we should let competitors collude and maybe that will be a better outcome. Far from it. But we should let companies use tools that will make them better competitors. And there is no doubt that these tools are efficiency-enhancing.

Could you say a bit about how AI (artificial intelligence) changes the picture, especially as we move into the agentic era in which AI agents will be making decisions autonomously?

That’s a much more interesting question. Not much has happened on that front yet, but we can see it coming. So far, these algorithms are decision-making helpers. They help companies make decisions about pricing or whatever.

With agentic AI, we will have technology that could allow companies to dispense with human decision-making. Section 1 of the Sherman Antitrust Act of 1890 — and antitrust law in most jurisdictions — focuses on conspiracies. That’s the evil the Sherman Act was designed to target. It was about busting trusts that made decisions collectively. With agentic AI, we’re moving into a world in which decisions are made at a much faster pace — in the blink of an eye — and it is virtually impossible to detect what AI decisions were based on.

We don’t know how antitrust law will treat this. I don’t know if there’s an enforcement mechanism that is viable in that context. When these new tools emerge, regulators and the law will have to adjust.

So there’s a lot to lot to learn as this space evolves.

For sure. And maybe we’ll need another AI tool to figure it out.

There you go. I appreciate you coming to talk to us today.

Thanks for having me.