When Computers Collude

NOTE: This is an excerpt of Planet Money's newsletter. You can sign up here.

If you shop online, there's a good chance the price you pay for stuff is determined by a computer algorithm. As of 2015, over one third of the 1,600 best-selling items sold on Amazon came from sellers who used algorithms to set their price. Algorithms are spreading like crazy, but are they giving companies too much power over consumers?

Emilio Calvano, an economist at the University of Bologna in Italy, has been studying the economic effects of algorithms. In 2016, he hopped on a scooter with his colleague Giacomo Calzolari and scooted across the historic city to their university's computer science department. There they conversed with experts on artificial intelligence and took a stack of materials to study.

A few years later, Calvano and Calzolari, together with a couple of other Italian economists, Vincenzo Denicolò and Sergio Pastorello, used what they learned to create a simulation. They built algorithms that use "reinforcement learning," which means they learn as they go through trial and error. "This is basically the building block of artificial intelligence," Calvano says. Their A.I. algorithms duked it out on a university computer, competing to sell imaginary goods in a virtual marketplace.

But these algorithms did not go to war and force each other to sell their goods at the lowest competitive price. Instead, the algorithms raised prices as if they were explicitly colluding together. The most surprising part is they did so even without being able to communicate with each other. "So in a sense they reach superhuman performance," Calvano says.

All of which challenges the rules — and even the entire vocabulary — of antitrust law. Is it even "collusion" if these machines aren't talking to each other?

Algorithmic competition: good or evil?

The use of algorithms by companies to set their prices is not new. The airline industry has been using them since the 1980s. However, those algorithms relied on close supervision by humans.

Today, algorithmic pricing is transforming a range of industries, from taxis to retail. In 2012, Amazon changed prices about 50,000 times a day. A year later, they implemented a dynamic pricing strategy — and they began changing prices over 2.5 million times a day. The strategy has pushed its competitors to do the same — and this isn't just businesses on the Internet. Brick-and-mortar retailers have been adopting electronic price tags so they can also flexibly change their prices.

Most economists believe flexible pricing is good for society, at least when it comes to overall economic efficiency and growth. When companies can tailor prices to people, more trades get done and more stuff gets made. While some consumers have to pay more, others get to pay less. Algorithmic pricing could also help prevent waste by, for example, lowering prices on produce as it gets closer to going bad.


Calvano says there are three schools of thought about how to regulate the use of these new algorithms.

The free-market camp believes it's not the government's job to tell companies how to set their prices, and that the market will ultimately fix itself. While companies are gaining more power with these superhuman tools, consumers will also increasingly be able to leverage algorithmic power to make purchasing decisions. Services like Groupon, LivingSocial, CamelCamelCamel, and Honey already exist to help you find bargains. If you use Kayak to buy airline tickets, there's a simple algorithm that helps you decide whether it's the right time to buy tickets. It's sort of a clunky feature, but it's likely that such algorithms will get better in the future.

A second camp, Calvano says, believes we should create a new regulatory authority that treats algorithms like drugs. It'd be sort of like the FDA, but for algorithms. They would test these algorithms in government computers before allowing them into the marketplace.

Calvano says he and his colleagues fit into a third camp. "Let's allow firms to innovate and deploy these algorithms in marketplaces because they can do a lot of good," he says. Then, he says, the government's antitrust authorities should monitor them and intervene if there's a case they're being used anti-competitively. Such an approach, however, might require modernizing our dusty antitrust laws.

Did you enjoy this newsletter? Well, it looks even better in your inbox! You can sign up here.

Copyright 2019 NPR. To see more, visit https://www.npr.org.