In the Journal of Medical Ethics, Joffe et al recently published an article titled “What do patients value in their hospital care? An empirical perspective on autonomy centred bioethics”.¹ This empirical study evaluates whether patients’ willingness to recommend their hospital to others is more strongly associated with their belief that they were treated with respect and dignity than with their belief that they had an adequate say in their treatment.* Joffe et al go on to suggest that confirmation of these empirical hypotheses would constitute a prescription for elevating the principle of respect for persons to the level that the principle of respect for autonomy currently enjoys in our model of the ideal patient–physician relationship (p 104).¹ In other words, they suggest that by some means empirical findings could influence our ranking of the normative principles. Earlier in the article, they make an even stronger claim about the influence of empirical data on our acceptance of normative principles. They suggest that, if it were demonstrated empirically that some patients prefer to delegate medical decisions to health care professionals, a serious challenge would be levied against the normative assumptions underlying the principle of respect for autonomy, at least under the mandatory autonomy view, which holds that patients not only have a right but also an obligation to act autonomously (p 103).¹ In the light of many recent empirical studies challenging the centrality of patient autonomy and shared decision making in bioethical theory, I think it is instructive to evaluate the means by which empirical findings, like those offered in Joffe et al, strengthen or weaken our arguments for ethical principles. In particular, I would be interested in how these authors propose that their data led them to the normative conclusions they reached.

In the last paragraph of their article, Joffe et al write: “we do not recommend that patients’ perspectives should unilaterally determine ethical frameworks. We do, however, believe that data such as those presented here can contribute to the search for reflective equilibrium in bioethics”(p 107). The term “reflective equilibrium”, as the authors note, was introduced by John Rawls. At least in its first instance, it refers to a way of constructing a moral theory by balancing one’s considered moral judgements against one’s moral principles, until one’s judgements and principles form a consistent set—that is, a moral theory (p 288).² Joffe et al’s idea seems to be that by surveying patients’ perspectives they will be able to capture one side of this equilibrium, considered moral judgements, or moral principles (they do not specify which), and in so doing contribute to the desired end: a consistent ethical framework to govern medical encounters, built (at least in part) from the principles and moral judgements of the patient community. Whatever the merits of this goal, however, Joffe et al fail to capture either the considered moral judgements or the moral principles of those they survey and so fail to contribute to the moral theory they seek to construct.

Rawls defines considered moral judgements as those judgements in which our moral capacities, which he considers analogous to our linguistic capacities, are “most likely to be displayed without distortion”—for example, those offered without hesitation, given without strong emotions like fear, and made in the absence of conflicting interests (p 47).³ The distinction between considered judgements and judgements generally is important. When constructing a moral theory for a particular community—for instance, the patient community—we want to use only those judgements that reflect the respondents’ real moral sensibilities, and not those stemming from superficial prejudices or their mood on the day they happen to respond. This raises two important questions, however, for researchers, who, like Joffe et al, are using the concept of reflective equilibrium: (1) precisely how considered do considered judgements have to be if they are to count; and, more practically, (2) how can a researcher know whether he or she is collecting them—that is, what survey method, if any, is appropriate for the task?

Although it is difficult to give a positive answer to these questions (and I will not attempt to do so here), some survey methods, such as the mailed questionnaires that Joffe et al used, seem particularly inadequate. Rawls suggests that certain external conditions favour the formation of considered judgements: “the person making the [considered moral] judgment is presumed … to have the ability, the opportunity and the desire to reach a correct decision (or at least, not the desire not to)” (p 48).³ Very likely, however, many of Joffe et al’s respondents lacked the necessary ability, opportunity, or desire to reflect on their moral judgements when responding to the questionnaire they received in the mail. Furthermore, even if a number of patients did offer legitimate considered judgements, there is no way to distinguish these from those made by respondents who lacked the requisite ability or desire. Although the size of Joffe et al’s study is of value for its ability more accurately to reflect a population’s response to its survey questions, because of the practical limitations that come with its size, the study falls short of capturing patients’ considered moral judgements.

Any empirical approach using reflective equilibrium, as Joffe et al’s, faces a second challenge: why do we want people’s considered moral judgements to influence our theories of ethics in the first place? In his influential critique of reflective equilibrium, D W Haslett writes:

… given the wide differences between people’s considered moral judgments, and given that these differences are, as we know, largely just a reflection of differences in upbringing, culture, religion, and so on, it would appear that, far from having a reason for giving people’s considered moral judgments initial credibility, we have instead a reason for initial skepticism (p 309).⁴

If moral judgements are liable to reflect superficial prejudices, one could argue, considered moral judgements are liable to reflect deep seated ones. Surely this prejudice is something ethicists would like to overcome, not codify. While I do no think this challenge is insurmountable,†⁵ it does demand that researchers justify the inclusion of considered judgements in ethical theory before using the method of reflective equilibrium. Joffe et al have failed to do this.

Joffe et al’s study is susceptible to a second line of critique. Even if the study’s use of mailed surveys is appropriate, it fails to capture either patients’ considered judgements or principles, because, put simply, it does not ask for considered judgements or principles. Instead, it asks patients whether providers respected their person or respected their autonomy, and then tests patients’ responses to these questions against whether they report being satisfied with their care. If a provider’s acting with respect for persons is a better predictor of patient satisfaction than him or her acting with respect for autonomy, Joffe et al conclude that the principle of respect for persons should be assigned as much importance, ethically speaking, as the principle of respect for autonomy. As should be clear, this conclusion does not follow from Rawls’s conception of how one constructs a moral theory. In a Rawlsian view,‡^3,6 a moral theory requires knowing which principles patients hold, not whether those principles are associated with patient satisfaction. Joffe et al seem to be operating with an underlying utilitarian assumption to the effect that what we ought to do ethically speaking is whatever will lead to the greatest patient satisfaction. Although there may be reasons for accepting this utilitarian assumption (which Joffe et al do not provide), certainly there are others for rejecting it. For instance, although patient satisfaction may give a hospital a very good reason to change a policy, we probably do not want to say this reason is a good ethical reason. It is just good business sense. This is an especially important point given the principles that Joffe et al evaluate. Respect for autonomy and respect for persons are traditionally viewed deontologically—that is, it terms of duties or rights, which are valued for their own sake rather than the consequences (such as patient satisfaction) that they produce. In any case, these utility considerations take us far from patients’ actual moral views, the very things Joffe et al, by invoking Rawls’s reflective equilibrium, propose to capture.

Lastly, there is a question of their instrument’s validity. As I have been arguing, Joffe et al claim to assess whether patients are treated according to the principles of respect for autonomy and respect for persons. Yet, their single item assessing respect for autonomy—the question, “do you feel you had your say?”—does not do the principle justice. The principle of autonomy not only requires that the health care provider asks the patient for his or her opinion, but also that the provider acts on the patient’s opinion. Their instruments are similarly inadequate for the principle of respect for persons, which, they suggest, includes “autonomy, fidelity, veracity, avoiding killing, and justice”, as well as “respect for the body, respect for family, respect for community, respect for culture, respect for the moral value (dignity of the individual), and respect for the personal narrative”(p 104).¹ How are we to know whether patients had all or any of these in mind when they answered the question: “Did you feel like you were treated with respect and dignity while you were in the hospital?” Joffe et al acknowledge that these ethical concepts are a bit unwieldy for a survey of manageable length. However, these practical considerations should be used not only to excuse the study but also to question its ability to clarify the ethical concepts it claims to assess. They should prod us to ask, regardless of the survey’s scale and the limitations that its size produces: does this survey really address what we mean by the principles of respect for autonomy and respect for persons?

With any empirical study in bioethics, there is a gap between the empirical hypotheses the study confirms and the normative conclusions its authors would like to draw from it. In their article Joffe et al hoped to bridge this gap by invoking Rawls’s notion of the reflective equilibrium. As I have explored, however, the study does not contribute to either side of the reflective equilibrium they imply, and, thus, they fail to demonstrate how their findings challenge the centrality of autonomy and shared decision making in bioethics.

Joffe et al’s failures are instructive, however, insofar as they suggest how we could better bridge the gap between research and theory. The use of the reflective equilibrium in empirical research has promise, provided researchers are clear about: (1) how to define considered moral judgements and/or principles; (2) how their methods capture these judgements and/or principles reliably; (3) how the inclusion of considered moral judgements strengthens rather than weakens bioethical theory; and (4) how their instruments are valid for the judgements or principles they mean to assess. In addition, empirical research can contribute to bioethics by questioning the assumptions implicit or explicit in our normative views. Joffe et al try to do just this when they argue in the introduction to their article (p 103)¹ that patients’ desire to delegate decision making challenges the mandatory autonomy view. However, if empirical findings are to defeat a particular normative principle, the assumption that those findings challenge must be logically necessary for our holding that principle. For instance, without showing that patients’ desire for autonomy is necessary for our holding the mandatory autonomy view, the studies that Joffe et al cite, even if valid, can be interpreted variously as devaluing the mandatory autonomy view or as recommending that we better educate patients on the value of autonomy. This normative question cannot be settled empirically.

Empirical researchers have the potential to contribute substantially to bioethics, but their work needs the kind of philosophical and empirical rigor that comes from truly interdisciplinary collaboration and must be informed by a careful reflection on the proper relationship between descriptive and normative ethics.⁷ Joffe et al take us part of the way down that path. An exciting research itinerary lies ahead.