Debates over information privacy are often framed as an inescapable conflict between competing interests: a lucrative or beneficial technology, as against privacy risks to consumers. Policy remedies traditionally take the rigid form of either a complete ban, no regulation, or an intermediate zone of modest notice and choice mechanisms.
We believe these approaches are unnecessarily constrained. There is often a spectrum of technology alternatives that trade off functionality and profit for consumer privacy. We term these alternatives “privacy substitutes,” and in this Essay we argue that public policy on information privacy issues can and should be a careful exercise in both selecting among, and providing incentives for, privacy substitutes.
I. Disconnected Policy and Computer Science Perspectives
Policy stakeholders frequently approach information privacy through a simpleinte balancing. Consumer privacy interests rest on one side of the scales, and commercial and social benefits sit atop the other. Where privacy substantially tips the balance, a practice warrants prohibition; where privacy is significantly outweighed, no restrictions are appropriate. When the scales near equipoise, practices merit some (questionably effective) measure of mandatory disclosure or consumer control.
Computer science researchers, however, have long recognized that technology can enable tradeoffs between privacy and other interests. For most areas of technology application, there exists a spectrum of possible designs that vary in their privacy and functionality characteristics. Cast in economic terms, technology enables a robust production-possibility frontier between privacy and profit, public benefit, and other values.
The precise contours of the production-possibility frontier vary by technology application area. In many areas, privacy substitutes afford a potential Pareto improvement relative to naïve or status quo designs. In some application areas, privacy substitutes even offer a strict Pareto improvement: privacy-preserving designs can provide the exact same functionality as intrusive alternatives. The following Subparts review example designs for web advertising, online identity, and transportation payment to illustrate how clever engineering can counterintuitively enable privacy tradeoffs.
A. Web Advertising
In the course of serving an advertisement, dozens of third-party websites may set or receive unique identifier cookies. The technical design is roughly akin to labeling a user’s web browser with a virtual barcode, then scanning the code with every page view. All advertising operations—from selecting which ad to display through billing—can then occur on advertising company backend services. Policymakers and privacy advocates have criticized this status quo approach as invasive since it incorporates collection of a user’s browsing history. Privacy researchers have responded with a wide range of technical designs for advertising functionality.
Frequent buyer programs provide a helpful analogy. Suppose a coffee shop offers a buy-ten-get-one-free promotion. One common approach would be for the shop to provide a swipe card that keeps track of a consumer’s purchases, and dispenses rewards as earned. An alternative approach would be to issue a punch card that records the consumer’s progress towards free coffee. The shop still operates its incentive program, but note that it no longer holds a record of precisely what was bought when; the punch card keeps track of the consumer’s behavior, and it only tells the shop what it needs to know. This latter implementation roughly parallels privacy substitutes in web advertising: common elements include storing a user’s online habits within the web browser itself, as well as selectively parceling out information derived from those habits.
Each design represents a point in the spectrum of possible tradeoffs between privacy—here, the information shared with advertising companies—and other commercial and public values. Moving from top to bottom, proposals become easier to deploy, faster in delivery, and more accurate in advertisement selection and reporting—in exchange for diminished privacy guarantees.
B. Online Identity
Centralized online identity management benefits consumers through both convenience and increased security. Popular implementations of these “single sign-on” or “federated identity” systems include a sharp privacy drawback, however: the identity provider learns about the consumer’s activities. By way of rough analogy: Imagine going to a bar, where the bouncer phones the state DMV to check the authenticity of your driver’s license. The bouncer gets confirmation of your identity, but the DMV learns where you are. Drawing on computer security research, Mozilla has deployed a privacy-preserving alternative, dubbed Persona. Through the use of cryptographic attestation, Persona provides centralized identity management without Mozilla learning the consumer’s online activity. In the bar analogy, instead of calling the DMV, the bouncer carefully checks the driver’s license for official and difficult-to-forge markings. The bouncer can still be sure of your identity, but the DMV does not learn of your drinking habits.
C. Transportation Payment
Transportation fare cards and toll tags commonly embed unique identifiers, facilitating intrusive tracking of a consumer’s movements. Intuitively, the alternative privacy-preserving design would be to store the consumer’s balance on the device, but this approach is vulnerable to cards being hacked for free transportation. An area of cryptography called “secure multiparty computation” provides a solution, allowing two parties to transact while only learning as much about each other as is strictly mathematically necessary to complete the transaction. A secure multiparty computation approach would enable the transportation provider to add reliably and deduct credits from a card or tag—without knowing the precise device or value stored.
II. Nonadoption of Privacy Substitutes
Technology organizations have rarely deployed privacy substitutes, despite their promise. A variety of factors have effectively undercut commercial implementation.
Engineering Conventions. Information technology design traditionally emphasizes principles including simplicity, readability, modifiability, maintainability, robustness, and data hygiene. More recently, overcollection has become a common practice—designers gather information wherever feasible, since it might be handy later. Privacy substitutes often turn these norms on their head. Consider, for example, “differential privacy” techniques for protecting information within a dataset. The notion is to intentionally introduce (tolerable) errors into data, a practice that cuts deeply against design intuition.
Information Asymmetries. Technology organizations may not understand the privacy properties of the systems they deploy. For example, participants in online advertising frequently claim that their practices are anonymous—despite substantial computer science research to the contrary. Firms may also lack the expertise to be aware of privacy substitutes; as the previous Part showed, privacy substitutes often challenge intuitions and assumptions about technical design.
Implementation and Switching Costs. The investments of labor, time, and capital associated with researching and deploying a privacy substitute may be significant. Startups may be particularly resource constrained, while mature firms face path-dependent switching costs owing to past engineering decisions.
Diminished Private Utility. Intrusive systems often outperform privacy substitutes (e.g., in speed, accuracy, and other aspects of functionality), in some cases resulting in higher private utility. Moreover, the potential for presently unknown future uses of data counsels in favor of overcollection wherever possible.
Inability to Internalize. In theory, consumers or business partners might compensate a firm for adopting privacy substitutes. In practice, however, internalizing the value of pro-privacy practices has proven challenging. Consumers are frequently unaware of the systems that they interact with, let alone the privacy properties of those systems; informing users sufficiently to exercise market pressure may be impracticable. Moreover, even if a sizeable share of consumers were aware, it may be prohibitively burdensome to differentiate those consumers who are willing and able to pay for privacy. And even if those users could be identified, it may not be feasible to transfer small amounts of capital from those consumers. As for business partners, they too may have information asymmetries and reflect (indirectly) lack of consumer pressure. Coordination failures compound the difficulty of monetizing privacy: without clear guidance on privacy best practices, users, businesses, and policymakers have no standard of conduct to which to request adherence.
Organizational Divides. To the extent technology firms do perceive pressure to adopt privacy substitutes, it is often from government relations, policymakers, and lawyers. In some industries the motivation will be another step removed, filtering through trade associations and lobbying groups. These nontechnical representatives often lack the expertise to propose privacy alternatives themselves or adequately solicit engineering input.
Competition Barriers. Some technology sectors reflect monopolistic or oligopolistic structures. Even if users and businesses demanded improved privacy, there may be little competitive pressure to respond.
III. Policy Prescriptions
Our lead recommendation for policymakers is straightforward: understand and encourage the use of privacy substitutes through ordinary regulatory practices. When approaching a consumer privacy problem, policymakers should begin by exploring not only the relevant privacy risks and competing values, but also the space of possible privacy substitutes and their associated tradeoffs. If policymakers are sufficiently certain that socially beneficial privacy substitutes exist, they should turn to conventional regulatory tools to incentivize deployment of those technologies. For example, a regulatory agency might provide an enforcement safe harbor to companies that deploy sufficiently rigorous privacy substitutes.
Policymakers should also target the market failures that lead to nonadoption of privacy substitutes. Engaging directly with industry engineers, for example, may overcome organizational divides and information asymmetries. Efforts at standardization of privacy substitutes may be particularly effective; information technology is often conducive to design sharing and reuse. We are skeptical of the efficacy of consumer education efforts, but informing business partners could alter incentives.
Finally, policymakers should press the envelope of privacy substitutes. Grants and competitions, for example, could drive research innovations in both academia and industry.
This brief Essay is intended to begin reshaping policy debates on information privacy from stark and unavoidable conflicts to creative and nuanced tradeoffs. Much more remains to be said: Can privacy substitutes also reconcile individual privacy with government intrusions (e.g., for law enforcement or intelligence)? How can policymakers recognize privacy substitute pseudoscience? We leave these and many more questions for another day, and part ways on this note: pundits often cavalierly posit that information technology has sounded the death knell for individual privacy. We could not disagree more. Information technology is poised to protect individual privacy—if policymakers get the incentives right.