A central interest in consumer law is the harm AI algorithms might cause to consumers. Firms are increasingly gaining the power to target individual consumers in manipulative ways and charge prices tailored to each buyer’s ability to pay. People may end up buying things they do not need or regret, at prices exceeding those in the pre-algorithmic market. Rivers of academic ink are spilled in describing the potential harms and recommending urgent regulatory action. Some of that work is very good, although the entire genre is suffering from an acute oversight: it ignores the documented benefits pricing algorithms are bringing to consumers. Personalized prices have been repeatedly shown in the empirical economic literature to benefit low-income consumers (and why not? The easiest thing for these algorithms to infer is individual purchasing power, calibrating the price to match it).
An assumption that runs through much of the legal literature on pricing algorithms is the passivity of consumers. Short of anonymizing themselves by changing the privacy settings (and good luck with that), there is nothing consumers can do to blur their profiling by sellers’ algorithms. Consumers, in other words, are price-takers, and are said to be in peril.
In a recent two-paper project, Haggai Porat challenges this view. Consumers, Porat shows, can fight back. According to Porat, the basic feature that consumers could exploit is the tendency of pricing algorithms to rely on the information conveyed by prior purchasing behavior. Specifically, algorithms assign a higher price to returning consumers who revealed a higher willingness to pay through their past purchases, and a lower price to other consumers who previously refused to buy. Consumers, in turn, if aware of this pattern, may strategically decline early purchases to secure lower prices in the future. In this way, consumers “bargain” with algorithms.
This cat-and-mouse price war between AI algorithms and feisty consumers can help some consumers enjoy lower prices, but it could also end up with other consumers paying higher prices. Overall, it could either enhance or reduce total consumer welfare. This is the lesson from Porat’s benchmark model. So outright bans on personalized pricing may not be smart. Mandated disclosure to consumers of the sellers’ algorithmic pricing practices is also a double-edge sword. Yes, it allows consumers to finagle lower prices; but unfortunately, it also triggers unpleasant counterstrategies by sellers, such as raising early-period prices.
Is it at all plausible to expect consumers to play this patient and calculated game of withholding early purchases only to secure better future prices? In a second leg the multi-paper inquiry, Porat offers novel empirical support, devising a clever laboratory experiment in which consumers make purchases over several rounds, with prices adjusting based on each consumer’s purchasing decisions in preceding rounds. In the lab, consumers were shown to indeed “bargain” with the algorithm, avoiding early purchases to secure better subsequent deals, and even more so if explicitly informed how the pricing algorithm works. Will they also do this with Uber, Amazon, and United Airlines?
It is perhaps premature to celebrate consumers’ arms-length bargaining power vis-à-vis AI algorithms, and Porat is endlessly cautious to use notions like “negotiation” and “bargain” only metaphorically. And yet, such inquiry could not be more timely. We are entering an era in which consumer-side AI agents are developed to assist people in maximizing individual preferences—in navigating the entire web to find low prices and suitable products. As these algorithmic consumers burgeon, longstanding goals of consumer law would become out of touch. Mandated disclosures would be redundant, dark patterns would pose no harm, and data privacy protection would be less necessary and might in fact backfire by handicapping these agents (who perform best when they know a lot about their human masters). Competition law’s burning worry of price collusion among sellers’ algorithms could, who knows, be offset by algorithmic consumer “cartels” coordinating to purchase at prices below sellers’ marginal costs. The deep-rooted conception of a “vulnerable” consumer would have to be categorically rethought.
