So far, debates on ethical aspects of data-driven medicine focus primarily on identifying and spelling out relevant principles of biomedical ethics in contexts of intensified datafication and computerized clinical decision support. However, there are actually a number of related but distinct pursuits that one might subsume under the heading of ‘ethics’—distinctions which have not yet been made explicit when reflecting on data-driven medicine.
As one example, consider Ross (1930), who in the course of advancing his realist, non-naturalist intuitionism about morality, distinguishes two domains of moral theorizing: the right and the good. Reflection on common-sense morality suggests that both spheres are indefinable, and in particular cannot be reduced to each other. Moreover, he puts forward a pluralist ethics which countenances not one, but a whole set of irreducible principles at the levels of the right and the good, respectively. This and his understanding of a prima facie duty have been influential for principlism (Beauchamp and Childress 2013, 15–16). In Ross’s terminology, rightness is a feature of acts, and rests on duties to fidelity, reparation, gratitude, promotion of a maximum of aggregate good, and non-maleficence (Ross 1930, 21). Goodness denotes intrinsic value, and is a feature of states of affairs, motives, and outcomes. Specifically, he lists virtue, pleasure, knowledge, and justice (Ross 1930, 134–41).
As another example, consider Habermas (1993), who draws a distinction between pragmatic, ethical, and moral employments of practical reason. In pragmatic employments, the will is fixed, and the focus is on techniques and strategies for implementation. In contrast, some decisions require reflection upon the will itself and what it takes to lead a good life: who we are, who we would like to become, and how we would like to live. In Habermas’ terminology, they pertain to the sphere of ethics. Yet another class of decisions transcends the perspective of the particular agent. The interests of others are taken into consideration impartially. For Habermas, a maxim is “unjust if its general observance is not equally good for all” (Habermas 1993, 7). Examining maxims in these ways is a matter of moral deliberation. It does not concern what is good for a particular agent, but raises the question of whether “a generally observed maxim is suitable to regulate our communal existence […], whether I can will that a maxim should be followed by everyone as a general law” (ibid.).
In the following, we seek to harness these framings while sidestepping the following issues as far as possible. First, we remain agnostic on the exact relation between the right and the good. For example, as can be seen from these brief expositions, Habermas seems to assume some interdependence between the good and the right, whereas Ross emphasizes their distinctness. What matters for our purposes is that both spheres can be distinguished and do not coincide. Second, we sidestep the question of how Ross’ and Habermas’ framings relate to each other, for example to what extent there is overlap between their notions of the good, and specifically to which sphere justice ultimately pertains. Ross acknowledges that the status of justice is complicated (Ross 1930, 26–27), but ultimately locates it, as mentioned, in the sphere of the good. In contrast, Habermas seems to locate justice in the sphere of the right, and we will accord with this in the following.
Setting these issues aside, let us note that extant debates on the ethical challenges in data-driven medicine leave open which of the foregoing spheres they address, i.e., whether rightness or goodness is at stake, and/or whether pragmatic, ethical, or moral considerations are called for. One initial proposal of ours is that in the context of data-driven medicine, the sharing of health data is in principle to be welcomed from the perspective of the good, i.e., bears the potential to realize intrinsic value, to enable fulfilled lives, and to put individuals in a position to promote the well-being of others.
As one example, availability of data opens up new avenues in biomedical research, public health, and clinical care by improving evidence bases, facilitating sound decision-making, and optimizing health outcomes. Computerized, algorithmic, and machine learning tools such as neural networks can be set up to recognize patterns in patient data, historical case data, chosen treatments, and outcomes, thereby highlighting patterns that would have otherwise escaped our attention. They can also be used as a safety mechanism to check proposed treatments for consistency with past choices. Independently of how such tools are put to work, sufficient amounts of data from the patient and her peers are needed to arrive at suitable reference classes, to train algorithmic applications, and to attune them to the envisioned task. For these reasons, generating data and making them accessible can in principle be welcomed from the perspective of the good. In Ross’ terms, we can note that such availability paves the way for knowledge generation and the optimized treatment of health conditions that inhibit pleasure. In Habermas’ terms, we can note that data-driven biomedical innovation can facilitate pursuits of our own notion of a fulfilled life.
Moreover, the needed data result from a bottom-up process of individuals sharing their personal data, ideally culminating in knowledge bases that strengthen the overall health of a population. In this sense, data-driven medicine, despite the individualistic spirit that underlies the idea of tailoring interventions to specific patients, is a context in which individual and collective good are intertwined (Sharon 2017, 100). The decision of individuals to share personal health data can be driven by a range of different motivations. On the one end of the spectrum, reasons of self-interest are conceivable. Such motivations are likely especially in settings where the prospect of benefitting from innovative tools is directly conditional upon contributing one’s data. The individual hopes to gain a better understanding of herself, her body, health, diseases, and risks of disease. Ideally, such knowledge can inform the choice of health services she receives and enhance their effectiveness.
On the other hand, sharing one’s data can be the result of motivations that do not rest on individualistic or particularistic goals, such as improvements of one’s own health. Sometimes, agents seek to transcend their individual sphere. Contributing to processes of scientific discovery and innovation by sharing one’s health data is one way to reinforce and to invest in the broader social context in which one exercises self-determination. It is not just that data sharing can be altruistic and beneficial. It can also be a way to participate in the pursuit of collective goals, and to play a part in constituting and reinforcing community. Indeed, gift theorists highlight that certain acts of giving are only fully understood if their aneconomic (Derrida 1992) aspects are taken into account. Gifts involve endowments, are not intended to prompt a return, do not merely advance self-interest, and convey symbolic, non-commodifiable aspects through recognition, dedication, and an investment of the donor herself into what is being given (Hénaff 2010, 2013). Such acts exceed a logic of exchange because they neither aim at a return nor could be offset against one. Gift theories make us aware that acts of giving can generate and reinforce social bonds, and open up new options in social space (Dabrock 2015, Braun 2017).
Specifically, acts of giving can be driven by solidarity. Prainsack and Buyx define solidarity as signifying “shared practices reflecting a collective commitment to carry ‘costs’ (financial, social, emotional, or otherwise) to assist others” (Prainsack and Buyx 2012, 346, italics removed). Unlike altruism and charity, solidarity is based on the agent “recogniz[ing] sameness or similarity in at least one relevant respect” (ibid., italics removed). In subsequent work, they argue that if an individual knowingly contributes data to a database that aims to create social value (as opposed to pursuing private sector interests), and accepts the costs of her act of sharing without the expectation to be rewarded, then her contribution is plausibly framed as an act of solidarity. Institutions that are transparent about how they use data, seek to generate social value in the first place, and do not offer inappropriate compensation to data sharing individuals are implementing solidarity-based governance (Prainsack and Buyx 2017, chap. 5).
Solidarity and justice share certain features, but also differ in crucial respects. In some sense, both concern the (re-)allocation of goods, but they differ with regards to the grounds that motivate these allocations. Justice alludes to fundamental and universal standards of what is owed to individuals, in a sense to be specified by one’s preferred substantive theory of justice. It is universal, gives rise to an unconditional ought, but does not require going beyond these foundational standards. Solidarity pertains to the (re-)distribution of goods that exceeds these minimal, universal standards. The degree to which we have reason to enact solidarity varies. We do not choose to enact the latter as a result of mandatoriness or even coercion, but because of further motivations not already implicated by justice. As indicated by Prainsack and Buyx, solidarity motivates this distribution by reference to distinctive forms of relatedness amongst individuals. Dabrock (2012, chap. 5) proposes to conceive of these social ties on the model of Ricœur’s (2005) gift-theoretic remarks on giving and gratitude: they are non-calculatory by virtue of not aiming at a return from the recipient, but resting upon a non-instrumental interest in fostering community. Ideally, this gives rise to gratitude in the recipient, who is then free to give back. Rather than “annul[ling] the gift” (Derrida 1992, 13) by introducing aspects of compensation, successive acts of giving can each retain an asymmetry between giver and recipient. Iterations of such acts and their recognition can deepen modes of integration amongst individuals. A sense of mutual obligation results, but one that has its roots in something exceptional rather than the universal, foundational norms of justice. In this way, acts of solidarity occupy a middle ground between justice and love, charity, or beneficence: not driven by a categorical ought, but also not just something extra, unowed to the non-intimate other.
Qualitative empirical evidence suggests that attitudes of solidarity and/or altruism might play a role in data-sharing behavior. For example, Facio and colleagues suggest that individuals who contribute their whole-genome sequencing data to research intend to help others, e.g., those at risk of genetic disorders, and to contribute towards genetics or health, “[b]ecause I can give back and be part of groundbreaking (and potentially life saving) research” (Facio et al. 2011, 1214). Similarly, Oliver et al. (2012) queried subjects about expected benefits of sharing their genomic data. Participants were anticipating to help future patients, in particular those with a similar condition, to advance medical knowledge, and contribute to general societal benefits realized by genomic research. In both antecedently reported data sharing preferences and actual data sharing decisions throughout the study, individuals showed a tendency to prioritize these benefits over individual privacy concerns. Mählmann et al. report that a majority of individuals in their sample were willing to share their genomic data, and that they were “driven by altruistic motivations of wanting to contribute to the greater good and accelerating research to improve the health of society” (Mählmann et al. 2017). Some individuals even share their genomic data without reference to a specific research project or purpose. Haeusermann et al. (2017) examine the attitudes of contributors to OpenSNP, a freely accessible database of genomic data published under a Creative Commons Zero license. While some of their motivations are primarily self-regarding, such as curiosity and the desire to learn more about their genome, other motivations indicate a willingness to contribute towards a common good, such as the desire to advance medical research and to improve genetic testing. A reoccurring theme throughout these studies is that while self-interested reasons certainly play a role in explanations of acts of sharing one’s personal health data, at least some of the motivational drivers go further.
There is room for additional research, for example on the following three questions. First, how pervasive are these other-regarding motivational drivers? For example, are they dominant across all populations, or do some demographics stand out? Second, a more fine-grained stratification of examined motivational drivers would be desirable. For example, to what extent are other-regarding motivations of data sharing indicative of—in Prainsack and Buyx’ terminology—solidarity rather than charity? That is, in what sense and to what degree do they rest upon recognition of reciprocity, sameness, and symmetry with potential beneficiaries of the envisioned research (Braun 2017)? Third, which safeguards and control mechanisms do individuals expect in order to facilitate exercises of other-regarding dispositions to share data? From the perspective of these individuals, what are desired and appropriate governance mechanisms?
For our purposes, we can highlight these questions as interesting, but largely sidestep the details of answering them. We do not claim that individual attitudes cohere with pictures of solidarity and gift-giving in the sense outlined above uniformly, consistently, and with statistical significance. We propose these pictures not as empirical claims about the motivations and attitudes of a majority of individuals, but as descriptive schemes to capture a set of target phenomena in all its complexity. Our claim is not that we must employ the descriptive schemes of gift-giving and solidarity when framing decisions to share personal health data, just that these schemes highlight attitudes, motivations, and intentions that might have otherwise escaped our attention. Neither do we claim that individuals must give, and that the importance of privacy should be deflated. Our observations are compatible with leaving it wholly up to the individual to reflect upon whether and how she seeks to contribute to the good of others, and if desired to withhold rather than to share data. The idea that sharing data is in principle to be welcomed from the perspective of the good does not preempt the question of whether data should be shared in a given case.
We have argued that an ethics of data-driven medicine should consider and make explicit whether and when it addresses the right versus the good, and how the sharing of data in the context of data-driven medicine can advance the good. Individuals’ pursuits of their notions of the good life sometimes take the form of giving and gifting. Data is a valuable resource that individuals can give, and sharing them can be intended as an act of solidarity and an attempt to contribute to a greater good.