Susan is Brad’s long-time oncologist. She has known him since he was diagnosed with cancer five years ago and has accompanied him through both chemotherapy and aftercare. Whereas the last check-ups were normal, the new examinations indicate that his cancer has returned. Having known Brad for a long time, Susan knows that he is very anxious, and that bad news often overwhelms or even panics him so that he cannot think clearly. However, she also knows that he is taking his alternative practitioner’s exam in three days and does not want to jeopardise his success. Brad has been studying for the exam for a long time and is looking forward to his new job. Even though Brad tries to appear confident and self-assured, Susan notices signs in his behaviour that indicate nervousness. As she is also aware of his exam anxiety, she decides to not tell him about her suspicions and invite him for a talk in a week.Footnote 1

This is a clear case of medical paternalism – at least, if we look at the medical ethics literature (Beauchamp and Childress 2019, 230–233; Groll 2014, 186–188; Buchanan 1978, 373). There are different interpretations of paternalism but withholding information to which a patient would be entitled to for the sake of their well-being is usually understood as a form of paternalism. While no decision of the patient is being overridden, his will is not being considered and he is deprived of the possibility to react to the information and then take action, if he wishes to do so. Generally speaking, “paternalism” means acting against or without considering the will of a person in order to promote their own good. Whether one’s will is autonomous or not is irrelevant according to some definitions of paternalism, but according to others, only interventions in autonomous actions and decisions can be considered paternalistic (Beauchamp 2010, 80–81). Either way, paternalism is usually seen in opposition to autonomy (Nys et al. 2007, 1).

From an individualistic understanding of autonomy, paternalism interferes with the independent sphere of an individual – a sphere in which they decide and act independently, and which must be protected from external interference. According to relational theories of autonomy, especially those that see relationships as constitutive for our autonomy, such an independent realm does not exist (Specker Sullivan 2016, 440; Stoljar 2022). They object to an understanding of autonomy in terms of ideal independence and self-sufficiency since we always act and decide within relationships and within a social network. What impact does it have on the understanding and evaluation of paternalism if autonomy is not perceived as an independent capacity, but as one that is in principle dependent on others?

Although there is now a large body of literature on both relational autonomy and paternalism, these two concepts have rarely been brought together. Especially in medical ethics, this can be considered an omission – for both a relational understanding of autonomy and the concept of paternalism are of concern in patient care. Against the backdrop of an individualistic understanding of autonomy, the paternalist (P) and the subject of paternalism (SP) are usually considered separately, as John Christman (2014, 371), one of the few authors to address the issue, notes. For example, the fact that Susan has known Brad for years is irrelevant to the evaluation of her behaviour. No attention is paid to the relationship between the two.

However, for the two other authors who address paternalism along with relational autonomy, this relationship is a decisive reason for evaluating Susan’s behaviour. According to Laura Specker Sullivan and Fay Niker (2018), conceiving of autonomy relationally opens the conceptual space that has so far been neglected in the paternalism debate: “when we assume a relational conception of autonomy, we can see that there is a type of intervention that, while resembling a paternalistic intervention, is importantly distinct” (650). With the concept of “maternalism”, the two authors aim to fill and analyse this conceptual gap. They want to clearly distinguish their concept from ‘classic paternalism’Footnote 2 and highlight that maternalism is to be considered under different justification criteria: the focus is on the relationship between the maternalist (M) and the subject of maternalism (SM). M must know SM well enough to judge that their intervention supports the autonomy of SM – and does not undermine it (655). According to the two authors, the added value of maternalism is mainly conceptual as interpersonal interventions in long-standing relationships of mutual trust can be conceptualised more appropriately (666).

There is no doubt that this is an important conceptual contribution to the paternalism debate. However, in the context of paternalism and relational autonomy, there are also urgent normative questions that need to be answered: What does it mean for the justification – perhaps also for the prevention – of medical paternalism if autonomy is understood relationally? Is Susan’s behaviour towards Brad justified over the same behaviour of a physician who has only known Brad for a few minutes? If this is so, then what does this mean for the type and quality of the physician-patient relationship that one should aim for?

In this paper I will analyse how the debate on medical paternalism can benefit from a relational perspective on autonomy. In the first section I will outline the concept of relational autonomy (1.). This will be helpful in view of the second section in which I will turn to the concept of maternalism. I will discuss the contributions of the concept to the debate on medical paternalism and relational autonomy but will also offer criticism (2.). In the last two sections, I will highlight how the understanding and justification of medical paternalism can benefit from a greater consideration of relational autonomy. First, I will discuss which aspects of the physician-patient relationship are relevant for the justification of medical paternalism against the background of a relational understanding of autonomy (3.). Based on this, I will then analyse how the physician-patient relationship and communication should be structured so that medical paternalism can be prevented. As will be shown, conceiving of autonomy relationally and acting accordingly means, above all, supporting others in their exercise of autonomy (4.).

1 Relational Autonomy

Most theories of autonomy, including those that look primarily at the individual, their inner life and reflexive capacities, include a relational component. What distinguishes the classical, ‘mainstream theories’ from the feminist-influenced, relational theories is the function attributed to this relational component.Footnote 3 In order to assess whether a conviction is autonomous, internalist theories look primarily at the way in which the conviction is acquired.Footnote 4 Following Harry Frankfurt’s hierarchical theory, for example, the question is whether the conviction has been reflected and accepted at a higher level of reflection (Frankfurt 1988). Even though internalist theories primarily look at internal processes, like the reflective endorsement of a desire, most of them do not completely disregard external influences (Friedman 2003, 87–91). For example, they see manipulation by others as an autonomy-inhibiting influence (Dworkin 1988, 18; Beauchamp and Childress 2019, 102). What is taken into account is the causal significance of external, social influences on autonomy; others can hinder or foster the development and exercise of autonomy. Some relational theories, on the other hand, go one step further: our social environment and our relationships not only influence our exercise of autonomy, but they are also constitutive for it.

Now, as Catriona Mackenzie and Natalie Stoljar (2000, 4) aptly point out, “relational autonomy” is an “umbrella term” under which various relational perspectives on autonomy are subsumed. What relational theories have in common, however, is that they reject an understanding of autonomy in terms of self-sufficiency and independence and take into account the social embeddedness of individuals. They do not consider autonomy in isolation from this basic human characteristic, but focus on the meaning of interpersonal relationships, social influences and circumstances on the exercise of autonomy, the acquisition of personal values and, more generally, the development of ‘one’s own self’. Not all of them consider interpersonal relationships and social circumstances as constitutive of autonomy (Westlund 2009; Stoljar 2022). However, in the following I would like to focus solely on constitutively relational theories, as they emphasise the importance of relationships for autonomy more strongly.

So, what does it mean that autonomy is constitutively relational? Very generally speaking, it means that certain relationships and social circumstances are an integral part of autonomy: I cannot exercise my autonomy without others, without my embeddedness in social contexts – even if I have the necessary reflexive capacities to do so. Of course, not all relationships and social circumstances are an integral part of autonomy, but only those that are conducive to autonomy. What this implies is defined differently by different theories: Examples are relationships in which I experience recognition by others (Anderson and Honneth 2005; Benson 1994; Mackenzie 2008) and a social environment that is free from oppression (Oshana 1998). If I experience recognition by others, am valued and taken seriously, I am more likely to perceive myself as an agent worthy of respect. If my environment is not oppressive, but gives me the opportunity to form my own opinions and develop my own values, this is more likely to enable me to lead an autonomous life. If, on the other hand, I do not receive the necessary recognition or live in oppression, I am conversely not autonomous – even if I otherwise fulfil all the prerequisites for autonomy. For, as mentioned, the relational aspects are considered constitutive for autonomy. Relational theories, thus, draw attention to the tension that social relations can, on the one hand, undermine our autonomy and, on the other hand, make it possible in the first place (Specker Sullivan and Niker 2018, 653–654; Friedman 2003, 86).

From an individualistic perspective on autonomy, Susan interferes with Brad’s individual sphere and restricts his autonomy. She deprives him of the possibility, to directly relate to the new information, to ask questions and to decide on the next steps on his own. From a constitutively relational view of autonomy, there is no sphere of ideal independence and self-sufficiency – we do not decide and act in isolation from others, but always in a network of social relationships and interactions. But does this mean that Susan is therefore not violating Brad’s autonomy? Or does she violate his autonomy in another way, for example, by not treating him as equal or not recognising him as an agent worthy of respect? According to Christman (2014), we can also diminish the autonomy of others through the attitude we adopt towards them. If a person is already vulnerable, we can further impair their autonomy by adopting an oppressive attitude towards them or by considering them incapable of deciding and acting for themself.

Susan could be accused of treating Brad as unequal because she does not currently trust him to make a rational decision about further therapies due to his nervousness about the exam. On the other hand, she enables him to complete his exams without worrying about his disease. Afterwards – hopefully with more peace of mind – Brad will be able to make a more autonomous decision regarding further therapies. Wouldn’t Brad want that himself? One could therefore also say that through her behaviour Susan does not undermine Brad’s autonomy, but enables it in two ways: with regard to his exams and with regard to upcoming therapy decisions. From a relational autonomy perspective, it seems therefore difficult to make a final judgement about Susan’s behaviour. Could a look at her relationship with Brad possibly be the deciding factor? That is what the concept of maternalism basically suggests.

2 The Concept of Maternalism

At first glance, the term maternalism may evoke associations of motherhood or mothering. The impression may arise that it refers to a more ‘feminine’ form of paternalism, perhaps a gentler, nurturing, more intimate or more empathetic version. In the context of nursing ethics, the term is in fact sometimes used in reference to the maternal role that nurses occasionally have towards their patients (Wright 2015; Wright and Hacking 2012).Footnote 5 Specker Sullivan and Niker (2018, 650), on the contrary, explicitly point out that they do not refer to behaviour that is usually classified as characteristically feminine. They just want to clearly distinguish their concept from the concept of paternalism – also terminologically.

I have already indicated that the concept of maternalism foregrounds the relationship between the intervening person (M) and the subject of the intervention (SM).Footnote 6 Also, I have suggested that Susan’s intervention may be classified as maternalistic. But what exactly is meant by maternalism? As a first step, I would like to answer this question. It allows one to better understand the contribution of the concept to the debate on medical paternalism as well as its shortcomings.

2.1 Definition

Maternalism: If paternalism is acting in another person’s best interests without due consideration of their autonomy, maternalism is acting for the benefit of another person in a way that takes that person’s autonomous agency into account, despite no explicit expression of consent or assent being given by the person on whose behalf the decision is made (655).

Based on this definition, maternalistic actions are – according to a classic paternalistic framework – “clearly wrong” (655) because they violate the individual’s right to make decisions in their legitimate sphere of control. However, there are two conditions that justify a maternalistic act according to the two authors. One is the so-called relational condition: M must be in a relationship of mutual trust with SM. The second one is the epistemic condition: M must know SM well enough to judge that their intervention does not undermine SM’s autonomy but supports it (655).Footnote 7 The two conditions indicate that Specker Sullivan and Niker presuppose a certain kind of relationship for maternalism: It must be a historical and ongoing relationship of mutual trust and understanding. For, according to the two authors, in a long-term relationship of mutual trust, it is possible to act on behalf of the other without communicating directly with them and seeking their consent. If we have known another person for a long time and have a trusting relationship with them, we usually have a certain amount of epistemic accessFootnote 8 to their pro-attitudes – including their “goals, character, values, priorities, and preferences” (657). And if we really care about that person, we have an interest in promoting their pro-attitudes and thereby support their autonomy (659–661).

It follows that also the motives underlying maternalistic acts differ from ‘classic paternalism’: M does not act without SM’s consent because they feel superior in their assessment of SM’s well-being, but because M knows SM so well that they can judge exactly what SM would want in a certain situation (Specker Sullivan 2016, 439; Peterson 2012, 3). Hence, M acts according to the currently unexpressed will of SM (Specker Sullivan and Niker 2018, 658).

According to Specker Sullivan and Niker, maternalism highlights the conceptual space for a kind of interpersonal intervention that is not covered by paternalism. If one follows a classical understanding of paternalism, the intervening person deliberately overrides the will of the subject or acts without asking for it. By focusing on the relationship between the intervening person and the subject and by reference to a relational understanding of autonomy, a third possibility opens up: out of a relationship of mutual trust and out of an interest in the autonomy of SM, M intervenes to support SM’s autonomous agency (660). To illustrate this briefly: From a paternalistic understanding, there is no difference between Susan’s behaviour and the behaviour of a physician who has only known Brad for half an hour and decides to withhold the diagnosis from him because he is not accompanied by a trusted person. The physician is convinced that it would be important for Brad’s well-being to hear about the news in the company of his partner. She does not know if Brad has a partner at all, or if he wants them to be there, but she acts on her conviction that it supports the well-being of all patients if they are not alone when breaking bad news. Both act paternalistically; both intervene in his individual sphere and withhold information from him to which he would be entitled. From a maternalistic perspective, however, there is a difference between the two cases: Susan acts against the background of her long-standing relationship and her familiarity with Brad. These provide her with epistemic access to his pro-attitudes, which the other physician does not have. This access will help her make decisions that correspond to Brad’s pro-attitudes, rather than her own or objective ones. But what does this mean for the evaluation and justification of Susan’s behaviour? Can her behaviour not also be unjustified? This question brings me to some critical remarks on maternalism.

2.2 Discussion

Let’s stay with the condition of epistemic access for right now. Specker Sullivan and Niker do point out that it is to be understood in light of the relationship between M and SM (658–659). However, I think they could make it even clearer what characterises this condition against the background of a relational understanding of autonomy. This is particularly important in order to understand the difference in the justification of classic paternalism and maternalism. Let us imagine, for example, that an assistant physician is Susan’s holiday replacement and has read through all the notes Susan recorded in Brad’s patient file. From the notes she knows about Brad’s nervousness before important decisions and that he has an exam coming up. From this alone she can conclude that it would be better, both in terms of his autonomy and well-being, not to inform him of the bad diagnosis until after the exam. This kind of epistemic access consequently does not necessarily presuppose a relational understanding of autonomy. It must therefore be emphasised even more clearly that the epistemic access itself has to be constituted by the relationship and grown historically. Thus, in contrast to the assistant physician, Susan knows Brad from the beginning of his therapy, she can better assess his reactions as well as aspects of his personality and has not only read about it. She has already experienced how nervousness manifests itself in Brad and can recognise it through certain (physical) signs. Therefore, it is not only the knowledge of certain facts about a patient that is relevant, but also how one acquires them and what experiences one shares with a patient. Access to certain types of information thus relies on real interaction and shared experience.Footnote 9 In section three, I will discuss in more detail this specific type of epistemic access and its significance for the justification of medical paternalism.

Closely related to the condition of epistemic access is a point of criticism that the two authors themselves take up: It is unclear what is meant by M having to know SM “sufficiently well” for the intervention to be justified, and how this is to be determined (662–663). According to Specker Sullivan and Niker, there is no direct solution to this problem as only after the intervention it becomes clear whether M knew SM sufficiently well, concretely whether the interests and autonomy competencies of SM have really been promoted by the intervention. But could this result not also be mere coincidence? As the example has shown, the assistant physician could come to the same conclusion as Susan and intervene in the same way – based on her knowledge of the patient file. Whether the outcome of her intervention could also be the same depends somewhat on what exactly is meant by the support of autonomous agency. And here I see another problem.

According to Specker Sullivan and Niker’s definition, maternalism “takes SMs autonomous agency into account” (655). M intervenes in order to support SM’s autonomy “relationally” (660). Even though the two authors add the qualifier “relationally” here, it is not quite clear what they mean by autonomy promotion and what kind of autonomy is to be promoted through maternalism (see e.g., 658, 662).Footnote 10 This could give the impression that, although maternalism is based on a relational understanding of autonomy, it promotes autonomy in individualistic terms such as independence and self-sufficiency. This perception may be seen as self-contradictory.

In the ‘classic paternalism debate’ there is also the idea that paternalism can be justified in favour of the autonomy of the subject: So-called “means paternalism”, for example, intervenes in the means the subject has chosen to achieve personal goals, but not in the goals themselves – on the contrary, the intervention is intended to serve the achievement of these personal goals (Dworkin 2020). And according to Dany Scoccia’s (2013) “loose paternalism”, P intervenes in the actions and decisions of SP in order to promote their idea of well-being. Besides, it is discussed whether “paternalism in the name of autonomy” (Sjöstrand et al. 2013), i.e., paternalism aiming at promoting and preserving the autonomy of the subject, can be justified. For if we consider of autonomy as something valuable in people’s lives, then it could also be justified to intervene in the actions and decisions of others in order to preserve and promote this value (Sjöstrand et al. 2013; Juth 2005, 101–102; Pugh 2020, 244–246; Caplan 2006 and 2008). Often, it is about protecting the subject’s future (global) autonomy through a short-term restriction of their current (local) autonomy (Caplan 2008, 1920).Footnote 11 For example, P feels it is justified to interfere with SP’s local autonomy in order to safeguard their (future) global autonomy, which refers to leading an autonomous life. However, even if most people value an autonomous life, SP does not necessarily have to share this viewpoint. Consequently, this ideal is imposed on them. This, of course, goes against the very idea of maternalism, as interventions should prioritize MS’s interests, regardless of whether it aligns with their global or future autonomy. So, it needs to be made clearer what it means to promote relational autonomy. I will go into more detail about what this can imply in the medical context in the remaining sections.

Even though I see the conceptual added value of maternalism, I will stick to the established term of (medical) paternalism in the following. In my opinion, sticking to the term makes it clear that the discussion about relational autonomy and paternalism has its place in the debate about medical paternalism. Besides, confusion with the term maternalism as used in nursing ethics can be avoided.

3 Relational Autonomy and the Justification of Medical Paternalism

To justify medical paternalism, it intuitively makes a difference whether a physician who has only known Brad for half an hour withholds information from him because of a general conviction that no patient should receive bad news without the support of a partner or friend. Alternatively, if Susan, Brad’s long-time physician, makes the decision to withhold the information, it presents a different scenario. This has to do with the relationship between Susan and Brad and her motivation to intervene, as the discussion of maternalism has shown. Nevertheless – and this is also suggested by Specker Sullivan and Niker (2018, 655, 661) – Susan’s behaviour is not entirely unproblematic, but requires justification. The reason does not stem from the violation of Brad’s individual sphere of autonomy, which is understood as ideal independence and self-sufficiency. This is because, according to a relational understanding of autonomy, such a sphere does not exist. Yet the action can be wrong in other ways. The crucial question then is how moral justification functions against this background. I will elaborate on it in the following in order to clarify which aspects of relational autonomy are relevant with regard to the evaluation and justification of medical paternalism, i.e., interfering in patients’ choices and decisions for their own good.

To recall: from an individualistic understanding of autonomy, Susan’s behaviour is morally problematic because she withholds information from Brad that are within his legitimate domain of control out of well-being considerations. Her intervention is justified if the benefit to Brad is so great that it outweighs the infringement of his autonomy – regardless of how well she knows Brad. If we now look at Susan’s behaviour from a relational understanding of autonomy, the situation is different; the fact that Susan intervenes in Brad’s individual sphere of decision-making is not in itself morally problematic. Rather, the evaluation of her behaviour depends on the extent to which the nature of the relationship with Brad and the resulting action are autonomy-constituting or not, and on the way in which autonomy is “shared” (Klein 2022).

So, what difference does it make to the justification of medical paternalism if we assume a relational understanding of autonomy? I would like to take up two aspects of relational autonomy that seem particularly relevant in this context: on the one hand the nature and quality of the relationship, and on the other hand, intersubjective recognition.

3.1 The Nature and Quality of the Relationship

As it has turned out, it is not the epistemic access to information about a person per se that determines the justification of an intervention from a relational autonomy perspective. A special kind of epistemic access must be involved, which is constituted by the relationship itself. Specker Sullivan and Niker address the character of the relationship they presuppose for a justified maternalistic act. They mention trust, mutual concern and understanding, which according to them is fostered by a long and intimate relationship. In order to elaborate on what it means to know another person well, they draw on criteria from Bonnie Talbert (Specker Sullivan and Niker 2018, 657):

  1. 1.

    We have had a significant number of second person face-to-face interactions with [person] A, at least some of which have been relatively recent.

  2. 2.

    The contexts of those interactions were such as to permit A to reveal important aspects of her/himself, and A has done so.

  3. 3.

    A has not deceived us about him/herself in important respects.

  4. 4.

    We have succeeded in accurately perceiving what A has revealed – i.e. we are not ‘blinded’ by our own biases or other impairments (Talbert 2015, 194).

The three authors agree that we gain knowledge about another person reciprocally – through a history of shared interactions (Talbert 2015, 203; Specker Sullivan and Niker 2018, 657). This is certainly an important aspect. I would like to further discuss the necessary requirements for a relationship to establish the specific kind of epistemic access that would justify an intervention based on a relational understanding of autonomy. These requirements increase the likelihood that P possesses the essential access to SP’s pro-attitudes, enabling interventions that promote SP’s autonomy and well-being interests rather than undermine them.

Following on from Talbert’s criteria, the first condition for this kind of relationship is continuity. In order to know a person really well, I need to be in continuous contact with them, for example, to see changes in their life plan or preferences. Of course, a certain duration of the relationship is also crucial, but continuity is even more important. Because even if I have known my sister since she was born and thus for many years, I cannot claim a special access to her current preferences if I only meet her for an afternoon every five years. With regard to Susan and Brad, we would therefore have to check how regularly Susan and Brad have met since the beginning of therapy five years ago.

Nevertheless, the duration of the relationship is not entirely irrelevant. For only if I know a person over a longer period of time will I also have experienced them in different phases of life. If I have been travelling with a person for five days and spend almost 24 hours a day with them, I have very close contact with them. Yet I have only ever experienced them in a ‘special situation’ – namely while travelling. It will be similar with many physician-patient relationships: Some physicians, especially those who work in the emergency department, will only experience patients in exceptional situations and will not know how they otherwise behave – in their everyday life, for example. Therefore, in addition to the length of the relationship, it is also crucial whether one has experienced the other person in various situations. Despite Susan’s long-standing acquaintance with Brad, a key question remains whether Susan has only encountered Brad in challenging decision-making situations or has had the opportunity to know him in different contexts as well.

But even if I have had continuous contact with another person over a long period of time and have also experienced them in different situations, this does not mean that I know them well and have a specific kind of epistemic access to their preferences. If I talk to colleagues exclusively about work and small talk topics, I do not really know them – even if I have seen them five times a week for several years and have also experienced them in different contexts (business meetings, Christmas parties, workshops, etc.). Consequently, the relationship must also be sufficiently meaningful that significant characteristics of the other person, such as their personality, values or preferences, become recognisable in conversations and other interactions.Footnote 12 Susan seems to know some of Brad’s character traits, but does she also know his preferences and what is important to him in life?

When we have a continuous and ongoing relationship with another person in which meaningful communication takes place, we get to know the person’s life history or narrative. Following theories of narrative identity, a person’s narrative is what constitutes their self, i.e. if we know the narrative of a person, we know who the person is (Pellauer and Dauenhauer 2022; Kühler 2020, 71): “Through narrative identity, people convey to themselves and to others who they are now, how they came to be, and where they think their lives may be going in the future” (McAdams and McLean 2013, 233).Footnote 13 I do not want to go into more detail on theories of personal identity or the self, nor do I want to explicitly advocate one of them.Footnote 14 However, I consider that the narrative approach is suitable to characterise the specific epistemic access more precisely: because we can understand a person’s narrative as an “epistemic tool” that helps us to make decisions on their behalf (Kühler 2020, 74, 82). It helps to understand the meaning and to assess the importance of certain decisions and situations in the person’s life.

If Susan knows Brad’s life narrative, then she not only knows that Brad is taking the alternative practitioner exam in a few days. She then also knows his career path so far, his life circumstances and his plans for the future and what passing the exam means against this background. She can then conclude that, against the background of his life history, it is better to withhold the diagnosis from him at this time. In a similar way, it is argued that the patient’s life story is a useful basis for proxy decision-making. A proxy who knows the patient’s narrative is able to continue it on behalf of the patient in the same way as the patient would have done themself if they were still autonomous (Blustein 1999; Rich 1997; Kuczewski 1994 and 1999).

Now, naturally, several objections can be raised against the aforementioned conditions of a relationship that constitutes the specific epistemic access. Firstly, it can again be objected that even knowledge of the life story does not presuppose a long-standing and intimate relationship. Thus, the assistant physician can know Brad’s life story just as well as Susan, if Susan has carefully recorded it in the documents. However, knowing the other person’s narrative is in itself not sufficient for the specific epistemic access, it is only one aspect of it. By knowing the life story, the assistant physician might be in a better position to decide in Brad’s favour, but she could not claim specific epistemic access to his pro-attitudes.

Another objection is that a person can change – especially in the face of drastic life circumstances such as a serious illness (Kühler 2020, 75, 81). Thus, their preferences, narrative identity and future plans can also change. If this happens suddenly, even a trusted person might not immediately learn of this change. They would then think they are acting in the spirit of the other person’s narrative, but they are not. For example, Brad might have decided not to take part in the exam only a few hours before the meeting with Susan – due to a sudden inspiration to devote himself entirely to his health. Susan would then not act in accordance with Brad’s current preferences, violating his autonomy and possibly Brad’s trust in her and thus their relationship. A paternalistic act can also fail in the case of ‘good initial conditions’, regardless of the underlying understanding of autonomy. So even though there is a longstanding and continuous relationship in which meaningful communication takes place and the life history is known, the intervention may not achieve the intended effect and contribute to SP’s autonomy and well-being – perhaps because SP has unexpectedly changed their preferences or P’s considerations were not careful enough. This is also the reason why paternalistic actions are prima facie morally problematic and always require careful consideration – whether they are based on an individualistic or a relational understanding of autonomy. However, I am concerned with naming criteria that, against the background of a relational understanding of autonomy, make it more likely that an intervention will also achieve the intended effect. These criteria should help to make a decision when it comes to the question of whether to intervene or not – the risk that the intervention will fail should thus at least be minimised.

I would like to mention another condition that constitutes the specific epistemic access. In order to be able to get to know meaningful characteristics and the life story of another person, they must present themselves to us in an undisguised way (Talbert 2015, 195). Susan can therefore only claim to know Brad’s preferences, personality traits and life story if he really reveals himself to her. This in turn presupposes – especially in a medical setting, which can be intimidating – that Brad feels he is taken seriously by Susan, trusts her and feels comfortable in her presence, which is an important prerequisite for shared autonomy in the physician-patient relationship. For Susan, this means that she has to show Brad to have a genuine interest in him, that she cares about him, his concerns and his fears. Susan must therefore treat Brad with concern and empathy. She must display an interest in supporting him in his intentions and plans. “Being supportive” is described by George Tsai (2018) as an important virtue in relationships.

These are unquestionably high demands on a relationship between physician and patient, which is actually a professional relationship. Whether it is justified to place such demands on the physician-patient relationship is something I will discuss in the fourth section. For the moment, I would like to summarise the nature of the relationship that constitutes the specific epistemic access that, from a relational autonomy perspective, is a relevant factor in the justification of paternalism: It is essentially a continuous relationship that has lasted for some time and until today, in which meaningful communication and interaction takes place, and which allows to get to know relevant preferences, character traits and the life story of the other person. In addition, the relationship is characterised by a sincere concern and interest in the other person, which allows them to show themselves without pretence and to trust the other person.

The access to SP’s pro-attitudes, the establishment of a trusting relationship, and a sincere interest in SP all contribute to increasing the likelihood that P’s intervention will strengthen SP’s interests. Another crucial condition, of course, is that P’s intervention is also motivated by these factors. I would now like to take up another aspect that plays a role in justifying paternalism against the background of a relational understanding of autonomy: intersubjective recognition.Footnote 15

3.2 Intersubjective Recognition

So far, the focus has been on the type of relationship that must be present for an intervention to have any prospect of justification against the background of a relational understanding of autonomy. Intersubjective recognition, on the other hand, is more concerned with the question of what it means to promote relational autonomy through paternalistic intervention. I mentioned that intervening in favour of autonomy is also discussed against the background of an individualistic understanding of autonomy. This usually involves an intervention to ensure that SP makes a truly autonomous decision. A well-known example in this context is Mill’s “bridge crosser”: a person who is about to cross a dilapidated bridge is stopped in order to inform them about the state of the bridge and thus enable them to make an autonomous decision (Mill 2011, 116; Scoccia 2013, 78–79; Nys et al. 2007, 13–14; Conly 2012, 18).

However, if we take a constitutively relational understanding of autonomy as a basis, then autonomy is not only to be promoted relationally – that is, through relationships – the autonomy itself is also to be understood relationally. So, what does that mean exactly? As I have shown, Specker Sullivan and Niker are vague about this. Overall, it is not really clear what they mean by autonomy promotion – apart from the fact that they mention that a successful maternalistic act also has positive effects on the relationship itself (Specker Sullivan and Niker 2018, 657–658, 661). As I also stated earlier, “relational autonomy” is an “umbrella term” that covers different relational understandings and theories of autonomy (Mackenzie and Stoljar 2000, 4). In the following, I will limit myself to one type of relational theories which are considered as ‘weak substantive’ accountsFootnote 16 and will show what exactly autonomy promotion through recognition means.

In the second section, I mentioned relational theories that consider recognition by others as an essential component of autonomy. These theories are sometimes referred to as “recognitional theories of autonomy” (Anderson and Honneth 2005; McLeod 2002; Mackenzie 2008; Stoljar 2022; Christman 2009, 181–186). While I cannot go into detail about individual theories, I would like to elaborate on their basic idea. Recognitional theories highlight how relational autonomy can be effectively promoted within the context of two-way relationships, such as the physician-patient relationship.

The core assumption of these theories is that certain attitudes towards oneself are necessary for autonomy and that these can only arise and exist within the framework of relationships that are conducive to these attitudes. Joel Anderson and Axel Honneth (2005), for example, regard the attitudes of self-esteem, self-confidence and self-respect as constituting autonomy – without these it is not possible to develop a sense of one’s own authority and to understand oneself as the author of one’s own life. Likewise, Paul Benson (1991, 1994) emphasises the importance of good self-esteem for autonomy: for an agent to be considered autonomous, they must feel worthy to act and recognise their own status as an agent worthy of respect, because only then can they develop confidence in their own competence to act. Carolyn McLeod (2002) looks at the importance of self-confidence for women’s autonomy in the context of reproductive medicine. Using the example of Anna, a patient who suffers a miscarriage, she illustrates that self-regarding attitudes such as self-confidence are essentially dependent on the behaviour of others, relationships with them and social influences. If Anna’s experience is dismissed as trivial by the health care professionals (HCPs) and she is left alone with her grief, this has a negative effect on her self-confidence. She will not feel that she is taken seriously and doubt her own experience. This reduces her autonomy in terms of overcoming the miscarriage, understanding its personal significance and also with regard to possible future reproductive decisions. If, on the other hand, Anna is shown understanding and is recognised as an equal agent, this has a positive effect on her self-confidence. According to McLeod, this imposes a moral duty on HCPs to enhance patients’ self-confidence as part of the duty to respect patient autonomy (56).

According to a relational understanding of autonomy, P encounters SP not only as a person who intervenes in their actions and decisions, but also as a person who expresses attitudes towards them and offers support to exercise their autonomy (Christman 2014, 371). Thus, Susan encounters Brad on the one hand as an agent who intervenes out of concern for him and on the other hand expresses certain attitudes towards him. At the end of the first section, I suggested that her behaviour could be interpreted as a lack of recognition of Brad’s normative authority: due to his exam anxiety and nervousness, she does not currently trust him to make a rational decision regarding further therapies. She could be accused of treating him as unequal and thereby manifesting the dominant physician-patient asymmetry of power, knowledge, and social status (Mackenzie 2008, 528). That may be an obvious interpretation.

However, if we assume that Susan does indeed have the specific epistemic access to Brad’s pro-attitudes discussed above and is also motivated by promoting his relational autonomy, there may be another attitude behind her behaviour: Susan knows the importance of the exam for Brad. Because of their longstanding familiarity and her knowledge of his life history, she is convinced that passing the exam will have a positive effect on his self-efficacy and self-confidence. This also applies to Brad’s confidence in being able to cope with upcoming challenges – such as a new outbreak of cancer. She is sure that the new job will have a positive effect on his desire to live and give him the energy to get through possible setbacks within therapy. Susan tries to imagine herself in Brad’s position; she too would be nervous before such an important exam and would not want to be disturbed by bad news in her final preparations. If we take all these considerations into account, we have to reconsider the judgement of Susan’s attitude: the fact that she is currently withholding the diagnosis from Brad can also be seen as an expression of respect for his authority and an attempt to strengthen his self-confidence and self-esteem against the background of his current situation.Footnote 17

However, if we understand autonomy as constitutively relational, Susan’s duty of autonomy is not exhausted in an appreciative attitude towards Brad. When it comes to the decision-making conversation after the exam, in which Susan informs Brad of the diagnosis and shows him possible therapy options, she must not only recognise him as an equal partner in the conversation and with regard to the decision, but also actively work towards a joint decision. The attitude she shows him must not only be one of recognition and mutual respect, but also one of invigoration and support – an attitude that gives him the feeling that she is standing by his side in a difficult situation and trying to find a decision with him that he can approve of. This is because the goal is not to empower Brad to make the decision by himself, but to make it within the relationship, which can strengthen not only his self-confidence, but also the confidence in the relationship. This is what is meant by not only promoting autonomy through relationships, but promoting relational autonomy through relationships. And if this motivation underlies a paternalistic intervention, then – against the background of a relational understanding of autonomy – this can be an argument for the justification of the intervention.

3.3 Summary

Based on the classic paternalism debate and an individualistic understanding of autonomy, the decisive factor for justifying medical paternalism is whether the benefit to the patient’s well-being outweighs the violation of autonomy rights inflicted by the intervention. If P’s motive is to protect or promote SP’s autonomy, as in Mill’s bridge crosser example, then the crucial question is whether the gain in (future) autonomy outweighs the current autonomy violation caused by the intervention. For the question of the justification of medical paternalism against the background of a relational understanding of autonomy, the following must be considered

  • whether the nature and quality of the relationship between P and SP is such that P can claim the necessary specific epistemic access to SP’s pro-attitudes, and

  • whether P’s motivation lies in nurturing SP’s pro-attitudes rather than their own, and

  • whether P expresses towards SP an attitude that is not only one of recognition and mutual respect, but also of interest in promoting relational autonomy.

As noted earlier, I have limited myself to a selection of relational theories. Therefore, it is conceivable that further justification criteria would emerge through the examination of additional theories.

4 How the Physician-Patient Relationship Matters

In the last section, I focused on the justification of medical paternalism against the background of a relational understanding of autonomy and worked out which aspects we have to take into account when deciding whether an intervention is justified or not. Now, one could argue that this work was almost in vain, since there will be very few physician-patient relationships that even come close to fulfilling the conditions of specific epistemic access, intersubjective recognition, and autonomy promotion. Thus, Specker Sullivan and Niker (2018, 663, 655) also point out that one should not fall prey to the deception that maternalism is easier to justify than paternalism. On the contrary, the conditions are very difficult to fulfil, maternalism is “quite normatively strict” and many maternalistic interventions must be considered unjustified.Footnote 18

This also applies to the conditions considered in the previous section. Very few physicians will have as intimate a relationship with their patients as Susan has with Brad. Very few of them will therefore have the specific epistemic access to their patients’ pro-attitudes. Even if some physicians wanted to get to know their patients that well, most of them would not have the chance to do so at all. Patients usually visit specialists only rarely, and encounters are brief. With the exception of GPs, most physicians do not accompany their patients over a long lifespan either. And these are only the purely practical problems. In addition, the provision and exchange of information is often limited to medical facts and little is known about the patient personally. This also raises the question of whether we would even want physicians to know so much about us. Besides, there is a certain potential for malpractice, which Specker Sullivan and Niker (2018, 663) also discuss. Physicians could simply pretend to have the necessary epistemic access to their patients in order to intervene in their personal interest. This could damage both the individual physician-patient relationship and trust in medicine in general.

With regard to the second justification condition, intersubjective recognition and autonomy promotion, one could argue that it would exceed the remit of the medical profession. Physicians usually have little time for each individual patient – how are they then also supposed to strengthen patients’ self-confidence and self-esteem? What should and should not fall within a physician’s remit is a discussion I cannot undertake here. However, I would like to point out that the answer to this question also depends on how much weight is given to patient autonomy in general and how many resources are made available accordingly to promote it. It is certainly not impossible to allocate more time and staff to foster patient autonomy more fully.

So, what follows from this? Against the background of a relational understanding of autonomy, do we have to consider almost all medical paternalism as unjustified? Is it then not simply a form of anti-paternalism? This is to be denied for two reasons. Firstly, it does not follow from the fact that medical paternalism will rarely be justified against the background of a relational understanding of autonomy that it is never justified (cf. on this also Specker Sullivan and Niker 2018, 660). As the example of Susan and Brad shows, relationships are conceivable in patient care that meet the necessary conditions. Secondly, the position of anti-paternalism is based on an individualistic understanding of autonomy. What anti-paternalism demands is opposed to a relational understanding of autonomy: According to this position, it is never justified to intervene in a person’s actions and decisions with a view to their well-being. The only reason to intervene is if there is doubt about a person’s autonomy (as in the “bridge crosser” example). But if autonomy is assured, the person is to be left alone with their decision (Christman 2014, Grill 2010 and 2012).

Nonetheless, the main benefit of the considerations regarding medical paternalism and relational autonomy is, in my opinion, to be sought elsewhere: namely in the contribution to the question of how the physician-patient relationship (and also physician-patient communication) should be structured so that medical paternalism is no longer even considered, because autonomy is actually shared.Footnote 19 From the criteria of specific epistemic access and intersubjective recognition, the following conditions for an autonomy-constituting physician-patient relationship can be derived:

1.) Physicians should consider the impact of their expressed attitude towards patients on fostering or undermining the patients’ ability to actively participate in therapy decisions. 2.) They also need to reflect on whether it makes sense regarding a decision to know more about a patient’s life narrative to be able to better understand the significance of the decision against this background. 3.) They have to meet their patients with interest and empathy so that they can build trust and show themselves without pretence. 4.) They should not consider themselves completely separate from their patients but perceive the relationship with them as an important factor in the treatment that needs to be protected. 5.) Difficult situations that may arise in the course of long-term treatment should be anticipated and discussed with patients in advance. For example, Susan could have discussed with Brad whether he always wants to be informed immediately about diagnoses, etc., or whether she should pay attention to certain life circumstances or involve friends and relatives in meetings. 6.) Patients should not be left alone with difficult decisions but should be accompanied in the decision-making process from beginning to end. Treatment decisions should be viewed as collaborative decisions rather than those made independently by two individuals.

I understand these criteria describe an ideal physician-patient relationship, the realisation of which is also hindered by structural problems in patient care. However, autonomy is such a central value in medical ethics – as in human life in general – that one should strive for everything that supports this value. And there are certainly efforts in this direction. The last criterion, for example, corresponds to the basic idea of a widespread understanding of shared decision-making (SDM): According to SDM physician and patient should meet as equal partners and reach a decision together. Some proponents of SMD appeal to a relational understanding of autonomy and emphasise the need to promote patient autonomy (van Nistelrooij et al. 2017; Dive 2017). With regard to SDM, however, there is still work to be done in order to really do justice to a relational understanding of autonomy (Lewis 2019; Gauthier-Mamaril 2022; Ubel et al. 2017; Bodegård et al. 2022).

5 Conclusion

Intuitively, it seems clear: for the justification of medical paternalism, it makes a difference whether a physician has only known a patient for a few minutes or whether she is familiar with them. In the classical discussion of paternalism, however, the relationship between P and SP is not considered as a decisive criterion for justification. Based on a relational understanding of autonomy, Specker Sullivan and Niker have made an important contribution to closing this conceptual gap with their concept of maternalism. With regard to medical paternalism, however, the normative question is of particular concern: What does it mean for the justification of medical paternalism if we take a relational understanding of autonomy as a basis? I have shown that the focus shifts from the question of the violation of individual autonomy rights to the nature and quality of the physician-patient relationship. It has become clear that medical paternalism can only be justified if physicians have the specific epistemic access to the pro-attitudes of their patients, meet them with an acknowledging and at the same time supportive attitude and actively contribute to a joint exercise of autonomy. As has been shown, only few physician-patient relationships meet these requirements, which is why medical paternalism must often be considered unjustified against this background. However, these criteria provide helpful indications of how physician-patient relationships should be structured in order to enable relational autonomy. There are already approaches that build on this idea, such as SDM – however, there is still much to be done to truly establish relational autonomy in patient care. Having outlined significant requirements, the next step is to delve into a more precise elaboration of what it entails in the physician-patient relationship.