THE APPEAL OF PRICE PERSONALIZATION

To begin with, consider “progressive pricing”:

An enticing argumentative move here is to use social welfare gains as a proxy for utility increases and then consider the fact that consumers indeed vary in their WTP. Take the influential interpretation of utilitarianism, the “dominant framework of welfare economics” (Coker and Izaret Reference Coker and Izaret2021, 391), on which persons and groups (e.g., members of firms) are morally required to maximize their contributions to social welfare.Footnote ³ On this view, a society with more total welfare is a morally better society.Footnote ⁴ Economists (e.g., Varian Reference Varian1985) have long argued that price personalization can increase social welfare given a utilitarian social welfare function, for instance by reducing or eliminating deadweight losses.Footnote ⁵ And the price personalization argument generally claims that, holding production costs constant, firms can increase social welfare by charging consumers with lower WTP less than consumers with higher WTP. To see why, consider Steinberg’s (Reference Steinberg2020, 102) helpful example in which a

Similarly, suppose Candace is affluent and Dale is not. If a journal charges Candace $35 for a subscription and Dale $5, this could increase social utility more than if the journal charged each person $20 (Coker and Izaret Reference Coker and Izaret2021, 390–91). If Candace’s additional $15 payment (paying $35 instead of $20) reduces her utility less than Dale’s savings of $15 (paying $5 instead of $20) increases his, then net social utility increases. If so, then utilitarians—who argue that our only ultimate moral obligation is to maximize total utility (again, often glossed as social welfare)—should endorse price personalization. A world with extensive price personalization is a higher-utility world.

Coker and Izaret (Reference Coker and Izaret2021, 392–93) also consider the moral value of price personalization through the lenses of other normative ethical views (e.g., egalitarianism). The authors stress that consumers can have higher WTP (a) when their income increases or (b) when they expect to derive more utility from an item.Footnote ⁶ The central point for our purposes is that pricing that is sensitive to individuals’ willingness to pay—and, therefore, not one-size-fits-all—is likely to enhance a society’s capacity to engage in welfare-increasing exchange.

As we begin to assess the promise and perils of price personalization, note that there is arguably no single dimension along which everything varies that a utilitarian might wish to measure or otherwise account for. Nonfinancial utilitarian goods, such as friendship, are not easily measurable or understandable, if at all, in terms of WTP or profit and loss. Here we get the “problem of noncommensurables” between financial and nonfinancial goods, where these different kinds of goods can equally support, or even constitute, human welfare. The following discussion is meant to be compatible with this caveat.Footnote ⁷

PRICE PERSONALIZATION: INITIAL OBJECTIONS

HOW MUCH MORALLY PERMITTED PRICE PERSONALIZATION COULD THERE BE?

My first counterargument to the idea that firms can reasonably price personalize concerns the claim that doing so would make a substantial positive difference in people’s lives. I contend that even if price personalization would increase social utility understood as social welfare, it will frequently be difficult or impossible to implement. Even if firms were legally allowed to price personalize, it is not at all clear how firms like Walmart and Amazon, which dominate our massive, anonymous and impersonal, markets, would know particular persons’ WTP.Footnote ¹⁴ AI-powered algorithms on Amazon, say, might help firms get a sense of WTP based on individuals’ long-term spending habits. But this would seem to be guess-work at best, for consumer preferences are hard and sometimes impossible to predict given that a consumer’s life is “an amalgam of numerous desires, beliefs, motivations, needs, interests, and preferences, all of which persistently change in real time” (Robson and Tosi Reference Robson and Tosi2025, n.p.). Worse for the proponent of price personalization, algorithms are not only limited in their upside potential in this regard but currently have no role in a vast number of market exchanges.Footnote ¹⁵

In our age of AI, firms face further moral and practical limits in potentially identifying consumers’ WTP. Thriving commercial societies have long embraced moral and legal protections of privacy, autonomy, and liberty. But while AIs are in some cases already gauging WTP algorithmically, they are frequently doing so without consumers’ knowledge and robust consent.Footnote ¹⁶ Three issues arise. One, it can be overly invasive for AIs to access, explore, use, and distribute sensitive consumer data (e.g., personal or financial data). Two, it relatedly can violate moral or legal privacy protections if AIs commandeer consumers’ rightful control of their data (Haenlein et al. Reference Haenlein, Huang and Kaplan2022), even if the AIs do not ultimately access and use the data. Three, millions of future consumers might respond to such worries about invasiveness and privacy by opting out of the data collection that makes WTP-based price personalization possible in the first place. Consumers might even pit AI against AI, using other AIs to identify and avoid AI-powered data trackers and aggregators. This last possibility could prove especially important for the utilitarian. Even if consumer opt-outs do not ultimately reduce utility-increasing exchange, firms’ resultant lack of data could indeed reduce consumers’ capacity to engage in it.

One might reply: “Well, price personalization can still increase utility in very many exchanges, even if far from all, so won’t a great deal of valuable price personalization still be possible?” Here is the problem. Consumers’ WTP often fails to track their objective good as human beings, understood here, arguendo, as utility maximization. Millions of consumers willingly pay high prices for tobacco, unhealthy food, and other health-reducing goods, yielding welfare losses. Price personalization can thus end up reducing, rather than increasing or not affecting, social welfare. It can support purchasing patterns that contribute to human languishing rather than flourishing (Robson Reference Robson2019; Haenlein et al. Reference Haenlein, Huang and Kaplan2022, 867).Footnote ¹⁷ Furthermore, while various participants in this debate (e.g., Coker and Izaret Reference Coker and Izaret2021) at times write as though we could take willingness-to-pay as a proxy for capacity-to-benefit, numerous studies in behavioral economics and psychology (e.g., Kahneman Reference Kahneman2011) show that the traditional “rational actor” model of consumers, when not positively misleading, is often inapplicable at best. Consumers are frequently un-, under-, mis-, or dis-informed, and act in ways that are suboptimally rational if not downright irrational. Charging wealthier customers more, or less wealthy customers less, could simply enable them to better satisfy malformed preferences. When satisfying preferences reduces consumer welfare, the problem of bad preferences becomes a problem for utilitarian defenders of welfare maximization.

To be sure, we should not advance an overly ambitious claim here.Footnote ¹⁸ The present claim is not that a large proportion of the items that consumers buy will be bad for them. It is that consumers fairly frequently buy items that are bad for them. In a given month, say, many consumers knowingly or unknowingly buy, for instance, unhealthy food or tobacco.Footnote ¹⁹ If consumers buy a hundred or more items per month, including food purchases, probably some of these will be bad for them. It is notable here that different consumer goods and consumption patterns are good or bad for different people. Tim might eat a meat lover’s pizza periodically, whereas Jill and Vincent eat two a week. Perhaps Tim’s health will be unaffected because of the infrequent consumption, Jill will benefit because she needs the protein for strength training, and Vincent, alas, will eventually have a heart attack. Here the problem for price personalization becomes acute insofar as utilitarianism enjoins firms not only to refrain from supporting consumption of bad goods but also to help consumers maximize their utility in their own, unique circumstances.Footnote ²⁰ This requires knowledge of WTP and particular circumstances that firms often do not have and cannot get (but see Robson, Reference Robsonforthcoming).Footnote ²¹

If price personalization is morally valuable insofar as it increases social welfare, there will still be situations where it cannot be carried out (and so has no value) or can be carried out but will reduce social welfare. In such situations, price personalization will be a wrong action or bad policy on utilitarianism, which treats outcomes alone as of ultimate ethical importance. On this ethical view, an action is wrong, or an institution, practice, or policy unjust or immoral (or, at least, less morally valuable), if it does not contribute maximally to utility, understood here as total social welfare.Footnote ²²

But even if price personalization often fails to increase utility, or its utility effects are unknown or even unknowable, perhaps targeted price personalization would increase utility. An account of targeted price personalization would need to determine the conditions in which price personalization improves lives and distinguish these from the conditions in which its net effect is neutral or negative. The burden is arguably on defenders of price personalization to provide such an account. Because absence of evidence is not evidence of absence, we cannot simply infer from the lack of such an account that no such account can be given.Footnote ²³ Whether defenders of price personalization should give such an account seems to depend largely on the extent and distribution of malformed (specifically, welfare-reducing) preferences in the marketplace. If such preferences are exceptional, perhaps the argumentative onus should instead be on proponents of uniform pricing to defend its value. Yet several scholars (see, e.g., Kahneman Reference Kahneman2011; Hirschfeld Reference Hirschfeld2018; Robson Reference Robson2019) have shown that malformed preferences are common in markets, driving a considerable degree of supply and demand behavior. Worse still for the proponent of price personalization, its effects are rather difficult to discern in actual markets, whose complexity and dynamism make successes difficult to identify.

PRICE PERSONALIZATION IN A DYNAMIC WORLD

Many accounts of price personalization focus on economic interactions that are static, nonstrategic, and one-off, rather than the dynamic, strategic, and iterative interactions in most actual markets.Footnote ²⁴ Other accounts consider scenarios in which consumers decide what to purchase at a time-slice, often in a micro-scenario regarding a single, particular purchase rather than a full pattern of consumption. Firms decide what to sell in this micro-case, and at which quantity and price. Cases such as these easily mislead. Markets are characterized by numerous, diverse, complex, and differentially temporally extended relationships between customers, firms, and other individuals and groups (Robson Reference Robson2023). Markets also involve multifarious patterns of iterative exchange rather than simple, one-off “spot” exchanges. As advertisers know, much of the value of these relationships is, for firms, a matter of perception. It is rather important, then, that “[b]y and large, consumers find price discrimination unfair, even when it advantages them” (Coker and Izaret Reference Coker and Izaret2021, 388; see also Huppertz et al. Reference Huppertz, Arenson and Evans1978 and Martins Reference Martins1995).Footnote ²⁵ Let us suppose that price-personalizing firms are taken to be unfair—whether or not they actually are (on unfairness, see Krugman Reference Krugman2000 and Moriarty Reference Moriarty2021; on fairness, see Marcoux Reference Marcoux2006 and Elegido Reference Elegido2011; Brennan and Jaworski Reference Brennan and Jaworski2016; Otteson Reference Otteson2019). In a market system with numerous interdependencies, perceived unfairness could have negative knock-on effects on consumer demand for not only price-personalized items but also other products within a firm’s brand. Consider the following example.

Suppose a consumer is upset that a mobile phone company price personalizes via an AI algorithm. Over the years, the consumer might not only not buy phones from the firm but also forgo buying computers that the firm sells as well, even if the computers themselves are sold with unitary rather than personalized pricing. Consumers who are morally repulsed by price personalization might even refuse to buy substitute goods elsewhere. This is bad news for the market-friendly utilitarian. Even if consumers resist price personalization based on mistaken imputations of unfairness, price personalization could end up reducing utility-increasing exchanges over the long term. To buy into the promise of price personalization within dynamic markets possessed of various interdependencies—as in our example of consumers’ interlinked preferences across diverse goods—we need an account with a reliable tally of the projected net results. Yet we are talking about markets with millions of firms in large and complex modern societies. That such a tally is unlikely to occur dampens the prospects of a robust defense of price personalization.

The effects of price personalization in a competitive marketplace will depend on not only particular facts about consumers but also particular aspects of firms and their competition. It is possible, for instance, that firms will call out price-personalizing competitors as unfair—as warranting customers’ and perhaps citizens’ “moral indignation” or “moral uproar” (cf. Steinberg Reference Steinberg2020, 102). For example, suppose that Coca-Cola encourages its retailers to price personalize the beverage, and many do.Footnote ²⁶ It is easy to imagine Pepsi or its retailers running counter-advertisements with messages to the effect that: “Coke is charging some people $1 and others $2! Is this any way to treat people?” Coke might respond that it is engaging in price personalization precisely for the good of all. What will be the net utility effect of these dueling firms?

The answer will depend largely on whether Pepsi vendors’ messages would reduce sales of Coke more than Coke vendors’ counter-messages would push sales in the other direction. Here, we have a case of competitive price personalization: not necessarily price personalization in perfect competition, but competition where one firm personalizes prices, the other does not, and they compete. If Pepsi wins the competition, and if satisfying one’s preferences reliably increases utility, even here, as traditional rational-actor economics usually says it will (more on this later), then there could well be a net utility loss—not despite but because of price personalization.Footnote ²⁷ The net result will also depend on how the utility loss from drinking a less preferred beverage compares with the utility gain from opposing apparent unfairness.Footnote ²⁸ A utility loss should come as no surprise. Opposing unfairness can lower consumers’ utility, as when one forgoes buying certain nice clothes in order to resist inhumane labor practices.

Of course, not all price discrimination will be called out. Some will go undetected. Among other reasons, firms can sometimes keep price discrimination secret. Universities in the United States, for instance, might already be engaging in price discrimination by offering students different amounts of need- and merit-based scholarships, partly based on their WTP.

Now suppose, finally, that in the moral realm, “ought implies can.” This long influential principle entails that any moral requirement to do x or refrain from x -ing requires that a person or group first be able to do so. If Jane ought to help the global poor, it must first be the case that Jane can help the poor. I have claimed that firms cannot price-personalize in very competitive markets without incurring large costs. It is not strictly true, however, that firms cannot. After all, firms can act in ways that predictably make them go out of business. Nonetheless, we do seem to get the following weaker claim: not “That a firm ought to x implies that a firm can x ,” but something more like: “That a firm ought to x implies that it is broadly feasible for a firm to x .” For example, while Walmart can in principle give all of its wealth to charity, it is also nontrivially true that Walmart cannot: If Walmart did, there would be no more Walmart. Giving all of its wealth to charity is “broadly infeasible” for Walmart.

The normative call for price personalization might not hold even on the weakened version of ought-implies-can. To see why, suppose Nigel has a 0.1% chance to x , and x is very morally important. Does this imply that Nigel ought to x ? That is unclear at best. The chance of success is vanishingly small. In the case of firms, price personalization that predictably would greatly lower firms’ long-term value, as the previous example of Coca Cola suggested it could, might well be irrational for firms to pursue on a standard expected-value calculation.Footnote ²⁹ It might cost substantial revenue, jobs, and future investment, lowering the long-term welfare of both firms and the societies in which they operate. The bar for what is feasible for a successful firm under the modified, less demanding ought-implies-can principle would then require keeping a firm functional and successful. Despite arguments (e.g., Coker and Izaret Reference Coker and Izaret2021, 391) to the effect that price personalization is morally required under, say, utilitarianism, it would then not only not be required, but, in fact, morally impermissible if price personalization would lower firms’ long-term value significantly. Whether it is impermissible in such a case will depend on many considerations, such as whether a firm’s eventual nonexistence, including any associated utility gains from the elimination of bad products, and any corresponding utility losses from the elimination of jobs, good products, and community value, would lower total long-term social utility. But in a competitive market, price personalization can be expected in many cases to decrease total utility, depending on the required resource outlay for personalization, any correspondence between consumer preferences and consumer good, and some additional considerations, which I will examine in more detail now.

Consider a final counterargument to the claim that price personalization will predictably increase, or even help to maximize, utility understood as social welfare: namely, that the rationality of price personalization will depend crucially on its opportunity cost (i.e., the value lost by forgoing the highest-valued alternative to price personalization), which will often be prohibitively high. If the risk-adjusted opportunity cost of x -ing is a loss of more utility than one gains (given the projected utilities of outcomes and their probabilities of obtaining), then it will be utilitarianly wrong to x . In their practical reasoning, the utilitarian might thus consider opportunity cost as rather important to their cost-benefit calculus. If buying x today would increase Samantha’s utility by (say) 300 points, she might also consider the alternative potential utility increase from, say, investing and growing that money for use 20 years hence. How, then, might firms increase social welfare (utility) by using resources they could have instead devoted to price personalization?

Suppose, for example, that price personalization would necessitate greater expenditure on AI software or labor that would help a firm identify data points such as consumers’ WTP. A rational firm must consider whether those resources could be used more profitably elsewhere, without simply assuming in this case that price personalization will be either cost effective or too costly. Many firms in the e-commerce world and beyond engage in price discrimination, judging it likely to increase profitability. Even so, firms that do not simply use fixed prices, but choose to personalize prices, must devote considerable resources to setting different prices for diverse consumer groups and perhaps even individual consumers. To do so, some firms might opt to buy AI, which could make personalization more feasible. Others might decide to pay for additional full-time employees or employee hours. Still others could choose to hire consultants. And still others might pursue a combination of these or other strategies. All such strategies take time and money. To understand the broadly utilitarian value or disvalue of price personalization, then, we must compare expected utility gains from it to expected utility losses, with a view to how firms might have increased social welfare by using their time, money, and other resources otherwise. Notice that this required comparison necessitates access to particular fact patterns that are often inaccessible to scholars and businesspeople alike, making categorical arguments for price personalization difficult or impossible to sustain.

One particular fact pattern that even firms deploying AI cannot ascertain seems to be individuals’ real-time willingness to trade off between goods. The related emerging literature on the limits of AI in central economic planning suggests that even AI will not enable price-personalizing firms to acquire such information about particular consumers (Robson and Tosi Reference Robson and Tosi2025; see also Hayek Reference Hayek2014 [1945]). Spurred to reflect at the point of purchase, customers make purchasing decisions largely intuitively. Although firms can increase social welfare by using “data and AI in ways that help rather than exploit customers” (Bergen et al. Reference Bergen, Dutta, Guszcza and Zbaracki2021, 8), consumers’ willingness to trade off across goods nonetheless changes often, hindering price-personalizing firms. Consumers persistently confront different bundles of purchasable goods, have changing budget constraints, and change their preferences relative to the same or new bundles.Footnote ³⁰ So even though AIs can indeed predict consumer preferences accurately at times, the background information required to predict well is often more complex than many realize. Not only do preferences and budgets change frequently; consumers’ major purchases (e.g., houses, cars) are also rare, and some one-off, making it harder still to predict consumers’ willingness to pay.

Before I conclude, it is worth indicating how the argument in this article also applies to second- and third-degree price discrimination. Second-degree price discrimination (or “versioning”; Steinberg Reference Steinberg2020, 99) involves charging different prices to consumers based on different quantities consumed, as when a firm charges less per unit for bulk purchases. Third-degree price discrimination (or “grouping”) consists of charging different prices to different demographic groups, as when students pay less for tickets to a university football game (Steinberg Reference Steinberg2020, 100; Twin Reference Twin2025). Like companies that engage in first-degree price discrimination, many firms that engage in second- and third-degree price discrimination will encounter large epistemic and moral challenges, including when they face the “ought-implies-can” requirement. Utilitarianly successful ventures in second- and third-degree price discrimination will require big data, practical wisdom and, at times, considerable luck. Consequently, it will often be difficult for firms to know just what people are willing to pay even when enhanced by AI-powered algorithms. Let me explain by noting a few general challenges and providing a concrete example.

On a spectrum of potential price personalization, social welfare will be maximized when firms engage in price discrimination somewhere between “0” and maximally. One way in which the three forms of price discrimination are similar is that attempts to identify the optimal level and kinds of price discrimination can backfire in cases ranging from trying to charge individual consumers what they are maximally willing to pay to offering diverse bulk discounts and discounts for students and seniors. As with first-degree price discrimination, the utilitarian moral status of these forms of price personalization will depend on how the resource reallocation it catalyzes affects the creation and redistribution of social surplus, for which consumers and producers compete at a time-slice. Here it is important to stress that economic surplus worth having in the utilitarian sense does not merely obtain when and because more preferences are satisfied, but only if additional preferences worth having (here, ones that increase utility) are satisfied. For example, if a firm offering a bulk discount gains less in producer surplus than they otherwise could have, and if consumers do not gain more than the producers lose, then the bulk discounts will be utilitarianly morally worse and, thus, impermissible. True, firms, policymakers, and scholars can perhaps know that bulk discounts on, say, tobacco will be utilitarianly worse over the long term because of the bad health effects across society of tobacco consumption. Still, many consumption items are not so obviously bad or good for people. In such cases, it may be acutely challenging to assess the likely utility effects of bulk discounts all-things-considered and over the long term.

As with consumers in utilitarianly justified price personalization regimes, the need to consume in ways that do not have a prohibitively high opportunity cost is important here. Suppose that farms offer grocery stores bulk discounts on organic food that is healthier than nonorganic food and costs more. It will usually not be obvious that even this price discrimination, rather than some other, will maximize utility. If the farms make less profit as a result, and thus enjoy less utility, we would need to know the utility effects of what they would have done with the additional profit they would have made. We would also need to compare it to consumers’ additional utility from the bulk sales, assuming there is any relative to the alternatives consumers face.Footnote ³¹ As with the above consideration of first-degree price discrimination, the comparisons here require access to particular fact patterns that are frequently inaccessible to businesses, scholars, and policymakers. At best, we do not know how and how far price discrimination would be utilitarianly valuable, if at all.

I do suspect, however, that third-degree price discrimination could have an especially positive effect on social welfare when pricing for fixed capacity. After all, the alternative here is more clearly disvaluable than alternatives to first-degree price discrimination (“price personalization”). The alternative is not charging a slightly different price, but leaving seats unsold (and, hence, unfilled) in a stadium, on an airplane, or in a movie theater, beds unsold in a hotel, and so on. Whereas price personalization generally focuses on how to charge particular consumers as much as possible without driving them away, third-degree price discrimination focuses on getting enough potential consumers to buy in the first place. It will often be easier to identify prices that a set of potential consumers are willing to pay than to identify prices that particular consumers are willing to pay.

To be sure, nothing crucial in my overall argument hangs on these reflections on second- and third-order price discrimination. But the relevant welfare effects are notable and, I believe, merit further investigation.