Regulatory angels and technology demons? Making sense of evolving realities in health data privacy for the digital age

How do we respect the legitimate privacy interests of individuals and communities about whom data relate while maximizing data’s utility as a fundamental resource in the 21^st century? Pyrrho, Cambraia and de Vasconcelos engage with this question, no doubt among the most enduring for the information economy in which we live and work. In their target article, “Privacy and Health Practices in the Digital Age,” the authors explore philosophical accounts of what privacy is—human right? a legal obligation? a social promise? or some combination?—and what it is not—a statistic, value-neutral, and a technical fix to a socially constructed problem.

I share the authors’ general frustration that health data privacy is seldom discussed as something other than sacrificial, as if protecting health information that is privately ours will somehow make us publicly sicker. Access to one’s own information is critical, in the authors’ view, to successfully rebel against opaque company practices that “produce and use aggregated data” about us but largely without our participation or consent. Pyrrho and colleagues argue that regulating the most prolific violators—big tech—is the best option we have for wrestling back some control over our private digital lives.

I comment on four of the authors’ positions in their target article. First, the authors’ tour of legal theories lacks mention of Helen Nissenbaum’s important contributions to constructing privacy as contextual integrity. I supplement the authors’ background by drawing on some of Nissenbaum’s work to better understand the social implications (intended and unintended) of distinguishing between the public and the private. Nissembaum’s ideas, and other scholarship on feminist approaches to privacy theory have shaped questions about who privacy is principally for, and also why some theorists believe only a privileged few can enjoy it.

Next, I explain why relaxing privacy and confidentiality rules during a public health crisis does not signal the State’s devaluation of privacy. Nor do episodic infringements on privacy in emergent times necessarily become the gateways to surveillance capitalism. Finally, I discuss why novel proposals to commodify data (sharing) have potential to introduce new incentive structures that enhance data agency and reinforce stakeholder choices in ways regulatory protections rarely have.

As actors in an increasingly data-driven world we iteratively appeal to dynamic choice architectures every time we decide what, where, and how information may be broadcasted about us. This idea underpins Helen Nissenbaum’s theory of privacy as contextual integrity (Nissenbaum 2004) and fosters relationship-building which Pyrrho and colleagues claim is part of “what privacy is for.” Nissembaum suggests that people participate in activities across different spheres and contexts which are each governed by distinct norms and practices. Privacy is therefore best understood, according to Nissembaum, as dynamic flows of information between contexts. One helpful privacy heuristic is whether emerging information technologies “violate any context-relative informational norms” (Nissenbaum 2004). Describing privacy as a set of dynamic choices as opposed to delimited frontiers between the “individual and the collective sphere” as Prryho and colleagues suggest, is more broadly inclusive of community value systems that reject such rigid frontiers exist at all. It can also help to explain how some frontiers work against marginalized groups and can in fact veil gender-based harm (Nash 2004).

Second, the authors cite an array of cases that implicate privacy—from social media use and artificial intelligence to genomics and election fraud. The diversity of their cases highlights the challenges some governments face in imposing adequate protections for data which sit at the intersection of two or more regulatory categories e.g. health data and consumer data. In this vein the authors propose that public health surveillance meant to guide the Covid-19 pandemic response demonstrated in part how State governments devalue privacy and inadvertently encourage its violation. However Pyrrho and colleagues level an unfair critique in this regard; their example conflates episodic infringements on privacy during emergencies with unfettered surveillance a la panopticon. Just as privacy protection supports strong “community formation,” infringing on privacy in exceptional circumstances may be necessary to keep those same communities safe. To generate the most public health benefits during a pandemic, health data needs to be protected not private. It is for this reason that privacy and confidentiality rules are relaxed during public health emergencies under the Health Insurance Portability and Accountability Act Privacy Rule, and the processing of certain categories of personal data to protect against serious cross-border threats to health is explicitly mentioned under Recital 46 of the General Data Protection Regulation in Europe, to name but two regulatory examples.

Another nuance the authors overlook in their arguments is that sharing data generates “common goods” only when those goods are actually distributed to the commons. Tech companies are not the only entities that fall short of such distributive justice targets. Studies demonstrate in my field of genomics, for example, that underrepresentation of diverse genomes has been one of many ways that precision medicine initiatives yield biased insights and skew bench-to-bedside translation (Ferryman and Pitcan 2018). Thus the authors’ appeal to accept frequent, though perhaps minor infringements on privacy to support health research is an exceedingly harder sell for those still waiting for basic benefits to trickle down e.g. access to quality healthcare (Tsosie et al. 2021). Recent efforts are promising to better align benefit sharing models with research participant values (Bedeker et al. 2022) and improve data representation (Wang et al. 2022), but have yet to measurably improve health outcomes for many patient populations so far.

Pyrrho et al. also criticize tech companies for converting “data into economic value,” because they have generated enormous profits seemingly at the expense of our privacy. But what if data contributors, not companies, were the beneficiaries of data sharing moving forward? This reality is no longer hypothetical. NFT (nonfungible token) marketplaces enable patients to share their health data for research or other purposes in return for crypto rewards (Kostick-Quenet et al. 2022). Blockchain and other distributed ledger technologies are unique in that they make data more secure by making it more accessible, thereby turning conventional understanding of data security and privacy assumed in the target article on its head. Health data sharing marketplaces that commodify both the sharing process and the data itself need not further disenfranchise communities as the authors imply. Rather NFT-based sharing could empower people through value-concordant data contribution and benefit sharing in real time.

Lastly, the authors present privacy as a parallel movement responding to “technological devices that threaten it.” In doing so, they inadequately credit the literal quantum leaps in computing that have accelerated innovation in privacy-enhancing technologies (see Aaronson (2013) for a highly entertaining history on the mathematical discoveries and privacy-relevant advances in quantum computing, and Perrier (2021) for the ethical implications of quantum’s modern applications in everyday life). Many data repositories entrust human oversight bodies to balance privacy as well as other autonomy-related values, for example, but increasingly rely on machines to computationally execute tasks that act on stakeholder values accurately, efficiently and at scale (Scheibner et al. 2021).

In presenting their lively account of the myriad people, purposes and policies served by privacy protections in the health research context and beyond, Pyrrho and colleagues convince readers of at least one truth: privacy isn’t one dimensional. The issues crystallized in this commentary compel new questions to add to the authors’ growing list published in their target article. These additional questions include but are not limited to who is privacy for, what metrics can be used to benchmark privacy loss or gain, and by what means can we protect privacy in line with diverse stakeholder values? To be sure, dialogue on these questions should be anything but private.