Harmful choices

When we know that someone is making a choice that will predictably risk or cause him to suffer significant harm, we have a basic duty of beneficence to try to prevent that.1 If James (box 1) was our friend or family member, we should try to discourage his choice and promote a better alternative. If we have a professional relationship with the person, (eg, if we are James’s doctor), we would have an additional professional responsibility to advise against this decision.1

But what if the person persists in his choice against our advice? Should we take further steps to restrict him? Should we restrain him or forbid him from making this choice? Should James’s employer insist on him being vaccinated to continue to work? Would a state be justified in mandating vaccination?

In terms of the law, and intervention by the state, one oft-cited response draws on an important liberal principle articulated by the philosopher John Stuart Mill. Mill’s ‘harm principle’ claims that the only justification for infringing the liberties of an individual is to prevent harm to others; harm to the self does not suffice. (Mill, p23)2 For debates about vaccination mandates in the COVID-19 pandemic, this means that ethical arguments have focused entirely on the effect of vaccination on risk of transmission to others, or on the use of scarce hospital resources. However, as the vaccination rate in the wider population has increased, and the pressure on hospitals has abated, we might return to the paternalistic reason for wanting James to be vaccinated.a Setting aside any question of harm to others, it would be far better for his own health for James to receive the vaccine.3 Could that possibly justify intervention to coerce or mandate vaccination?

The harm principle provides a simple and firm bolster against paternalism, against others’ well-meaning interference in our own lives. But it draws a sharp line between decisions that are harmful to the self and those that are harmful to others. It seems to imply that these are radically different types of decision and demand different ethical responses.

In this paper, I explore some reasons suggested by the philosopher Derek Parfit for dissolving or blurring the distinction between harm to self and harm to others.4 On Parfit’s account, some apparently self-harming decisions are relevantly like harming someone else.b He noted that his reductionist account of personal identity might be considerably more permissive of paternalism than traditional ethical approaches, though did not clearly identify if paternalism would be justified in cases like that of James. In the paper, I identify two different versions of reductionist paternalism, according to which the harm principle is undermined and health professionals and states may be justified in being paternalistic in a wider range of cases. The reductionist might claim that paternalism is more easily justified, or alternatively that what is conventionally thought of as hard paternalism is not actually ‘paternalistic’. Although I will take the first approach in this paper, the second would be equally plausible.

I also explore the relevance of Parfit’s ‘wide value-based objective view’ of reasons and suggest that this supports what I call Identity Relative Paternalism. To my knowledge, this paper is the first to draw a connection between Parfit’s later writing on the nature of reasons (his wide-value-based objective view), and his earlier work on personal identity. I will focus on hard paternalism, since it is here that the reductionist account of identity may have most radical implications,c and will not discuss specifically other arguments in favour of paternalism.d

I will focus here on medical examples and refusal of treatment by patients, identifying a spectrum of cases where relations of psychological connectedness and continuity might hold to stronger or weaker degrees despite intact cognition and in the absence of brain damage. One reason for focusing on patient refusals of treatment is that these typically (vaccination is an exception) are interpreted as being associated with harm only to the individual and not to other people. In cases where patients demand treatment (that the doctor believes will cause harm to the patient and is tempted to paternalistically refuse to provide) there may often be implications for resources that could potentially have impacts on other patients and provide a separate justification for refusal.e

Much has been written previously on cases involving significant changes in personality and cognition—particularly in relation to the validity of advance directives in dementia.5–9 For this paper, I will briefly outline some possible implications of Identity Relative Paternalism for these cases in the section on practical implications, but will not discuss dementia cases in detail. That is partly because I am interested in exploring the wider potential implications of a reductionist account of identity for paternalism (ie, beyond cases of brain injury or dementia). Second, questions about advance decisions and dementia are complicated by potential changes in moral status and lack of capacity in the later time points, so the individual’s contemporaneous views or interests might or might not be taken to have lesser ethical weight than the prior individual’s.

The harm principle

the only purpose for which power can be rightfully exercised over any member of a civilized community, against his will, is to prevent harm to others. His own good is not a sufficient warrant (Mill, p23)2

Mill articulated his harm principle early in his work, ‘On Liberty’. Mill was writing specifically about the justification for government actions and in particular the important limits on government interference in individuals’ lives. His harm principle warns against enacting paternalistic laws—ones that are designed to restrict individuals’ choices for their own good. The same principle, though, can be extended to other situations—for example to medicine. It provides a strong rebuttal against doctors who might be inclined to make paternalistic decisions motivated by concern for patients’ well-being. According to the harm principle, doctors should not limit patients’ freedoms for the sake of the patient’s own good, though they may be justified in doing so if the patient’s actions would cause harms to other people.f

Mill was forthright about the absolute freedom of individuals to make decisions that would affect only themselves: ‘Over himself, over his own body and mind, the individual is sovereign’. (Mill, p22)2 ,g In contrast, ‘[a]cts injurious to others require a totally different treatment’. (Mill, p140)2 ‘The distinction between the loss of consideration which a person may rightly incur by defect of prudence or of personal dignity, and the reprobation which is due to him for an offence against the rights of others…makes a vast difference both in our feelings and our conduct towards him’. (Mill, p141–2)2

Mill provided two key arguments against paternalism. The first was based on the value for individuals of freedom to have opinions and to act. Such freedom is instrumentally valuable because it enriches individual lives and allows the development of human faculties. ‘The free development of individuality is one of the leading essentials of well-being’. (Mill, p102)2 The second was on the basis of fallibility of paternalistic judgements. Such judgements are necessarily based on presumptions that may be mistaken. (Mill, p1150)2 The individual’s own self-knowledge is much more reliable.h I will return to these arguments against paternalism.

Harmful decisions

To help identify the relevance of these ethical arguments it will be useful to have some practical examples. I have mentioned already the case of James and his vaccination refusal. Box 2 and box 3 illustrate two more.

These three cases (boxes 1–3) involve adults making decisions in relation to their own life. For the sake of argument, I will assume that at the time of the decisions, each of these patients had capacity—that is to say that they had no disorders affecting their thinking or ability to reason. If formally assessed, it would be clear in each case that they were able to understand and retain the relevant information provided to them, weigh the reasons and communicate their choice.10

In these cases, many might be inclined to question the rationality of the decisions made. The risks that James, Jenny and John are taking with their own health appear considerable and the reasons that they cite would appear (to most people) to be insufficient. Health professionals might try hard to change the patients’ mind. They might even seek a formal psychiatric evaluation. Nevertheless, it is likely that the decisions would ultimately be respected.⁹ After all, that is what respect for the harm principle and for patient autonomy is thought to require.

By way of contrast, we could imagine the following cases (boxes 4–6).

In these versions of the cases it seems clear that the third party decision makers would not and should not be permitted to make these harmful choices.¹⁰ The mother’s refusal of a vaccine for James should be overruled. The judge’s decision should be urgently appealed. Someone else present when John collapses should call the ambulance against Michael’s wishes. It is clear in the third-party versions of these cases that even if these decision-makers are autonomous adults, their decisions will cause harm to other people, and on that basis, should be overruled.

But is there such a sharp divide between the first-person and third-person versions of these cases?¹¹

Identity and paternalism

In his book, Reasons and Persons, the Oxford philosopher Parfit famously defended what he called a ‘reductionist view’ about personal identity. (Parfit, p199–345)4 Parfit argued that the continued existence of an individual over time can be reduced to certain physical or psychological continuities. On this view, we can identify whether and to what extent someone at time point t1 is physically and/or psychologically continuous and connected with a person at a later time point t2. For example, we can ask ‘do they share the same body, the same memories, the same patterns of thought and character traits’? According to Parfit, the answers to these questions will tell us how the earlier person is related to the later person, and what is more will tell us everything that matters.¹² Although we might be tempted to ask ‘but is it the same person at t1 and at t2?’, according to Parfit, this question is sometimes empty. (Parfit, p213–4)4 At least in some special cases, there is no separate answer to this question. Once we have identified the relevant connections there is nothing additional to meaningfully say.

Should we be reductionists about personal identity? Parfit’s argument for reductionism is complex,12 and based on thought experiments involving divided brains and teletransportation. It is outside the scope of this paper to fully outline these arguments or to defend them in detail. However, it may be helpful to set out a brief intuitive case in favour. When someone has a profound brain injury and permanently loses the capacity for consciousness, family members often report that the person who they were has ‘gone’.13 There is a strong sense in such cases that even if the individual’s body remains alive, that the loss of psychological capacities is the end of their existence. In other cases, brain disorders can lead to profound changes in behaviour and personality. For example, in 2003, a 40-year-old school teacher was found to have an egg-sized tumour in his right frontal lobe.14 The teacher (having never previously behaved in such a way) had developed progressive uncontrollable sexual urges including paedophilia over a period of 3 years, and was diagnosed with a brain tumour only the day before planned prison sentencing. His sexual urges abated after resection of the tumour. What should we say about such a case? It seems highly plausible that the individual who made sexual advances to his prepubescent daughter was different in an important way from his earlier (and then postresection) self. That seems highly relevant to an understanding of whether we should hold the teacher responsible for the past behaviour. In these two cases, there might be disagreement about whether the person still exists after the brain injury, whether it is the same person with or without the brain tumour. But one thing that we should agree on is that the personality loss or change matters profoundly.

The reductionist view about personal identity has a number of striking implications¹³. One is that rather than thinking of oneself as a single self, existing from birth to death, it may be useful and natural to conceive of having ‘successive selves’. (Parfit, p305,19)4 Literature and colloquial language sometimes refers to ‘an earlier self’ or ‘a later self’. In usual circumstances, there will not be clear boundaries between these, but they are nevertheless distinct in important ways.

A second implication is that our concern about our own future (our egoistic concern) may not be binary (all or nothing)—instead it is a matter of degree.¹⁴ So, the person A at time point t1 can be closer to the later person B at t2, or further away—depending on the extent of psychological connectedness. The relevant question to ask is—‘how closely is A related to B’?, rather than ‘Is A identical to B’?

There are also a number of moral implications of this view about personal identity. One is that the boundary between the self and others is less distinct and less important. (Parfit, p338)4 Conventionally, decisions that affect only ourselves are taken to be outside the scope of morality. (Parfit, p319)4 It might be unwise, or even irrational to make a decision that will cause future harm to ourself, but (on a standard liberal view) it is not a question of morality. (Feinberg, 56)15 Parfit rejected that. He claimed that because the future self is relevantly like a different person, we should think of decisions that affect them in the same way that we think about decisions that will affect a different person. This gives rise to the Parfitian claim about what we morally ought to do

Reductionist Identity Moral Claim: “If we now care little about ourselves in the further future, our future selves are like future generations…Like future generations, future selves have no vote, so their interests need to be specially protected… We ought not to do to our future selves what it would be wrong to do to other people”,(Parfit, p319–20)4

In a very short section of his book, Parfit explicitly extended this to defend paternalism. He claimed that coercion or infringement of someone’s autonomy could be justified to prevent the individual from causing great harm to himself for no good reason. While we cannot justify restricting someone’s personal freedom on the grounds that they are acting irrationally, this is justified if they are acting wrongly. Parfit claimed that individual autonomy does not outweigh such moral concerns:

‘Autonomy does not include the right to impose on oneself, for no good reason, great harm’. (Parfit, p321)4 He went on, restating the moral claim in terms of the obligations of other people to prevent imprudence

Reductionist Identity Paternalism Claim: “We ought to prevent anyone from doing to his future self what it would be wrong to do to other people” (Parfit, p321)4 ,¹⁵

Parfit briefly acknowledged two standard objections to paternalism—that it is good for people to be able to learn from their mistakes, and that generally the individual will be in a better position than others to know what is best for him or her. These objections (which we could call the consequentialist and epistemic objections) are closely related to the arguments given by Mill and noted above. But in Reasons and Persons, Parfit did not clarify whether he thought that these objections outweighed the arguments that he set out in favour of paternalism given a reductionist account of identity. He provided no clear answer to the cases of James, Jenny and John.

Identity and treatment refusal

One point to note about the consequentialist and epistemic objections to paternalism is that they may not always apply. For example, there is little personal learning possible from a fatal error. (That would appear to potentially apply to the treatment refusal cases outlined above.) There will also be cases where we can be confident that the individual is mistaken about their own interests. James, Jenny and John in the treatment refusal cases are making serious errors of judgement. If we respect their decisions, that is not because we think they might be correct, but rather because they have a right to act imprudently. Furthermore, when we compare the first person with the third person versions of the cases, it is clear that neither the consequentialist nor the epistemic objections would come close to justifying third party harm. No matter how strongly we support their freedom to develop and have their own opinion or their knowledge of the individuals concerned, we would not allow Mary, Michelle or Michael to harm others seriously as a consequence.

Does the Reductionist Identity Paternalism claim apply to treatment refusal? There are two possibilities. One is that this claim applies equally to all instances of self-harm. It would apply equally to decisions that individuals make that affect them in the very near future, as it does to decisions that individuals make that affect them in the distant future. We could call this

Time-neutral paternalism: Individuals should be prevented from doing to themselves (whether in the near or in the further future) what it would be wrong for them to do to others.

However, as noted, Parfit’s view was that what matters morally is a function of psychological continuity and connectedness over time. This might suggest an alternative version.

Identity relative paternalism: Individuals should be prevented from doing to future selves (where there are weakened prudential unity relations between the current and future self) what it would be wrong for them to do to others.

On this identity relative account, our response to cases may vary—depending on the relationship between the person making the decision and the later-self harmed by it.

For example, in the three cases described at the start of the paper there appear to be different degrees of connection between the individuals making decisions, and their future selves harmed by their decision. In Vaccine refusal, there is likely a short period of time between when James_t1 makes a decision and James_t2 potentially comes to harm because of contracting severe COVID-19. This means that there are strong psychological connections between the different Jameses. In contrast, in Advance resuscitation refusal, many years have passed since John_t1 made his somewhat rash advance treatment decision. Although John_t2 has psychological connections with his previous self, we might suspect that they are somewhat weak. It was an earlier self (in his transient Dice Man phase) who made the advance decision to refuse treatment. John_t2 is likely to have many different interests, tastes, preferences and priorities from John_t1. The things that were important to him then are likely to be much less so now. Hospital birth refusal lies somewhat in between. In this case, there is a short temporal distance between the pregnant Jenny_t1 and the later Jenny_t2 who would potentially suffer a catastrophic complication of childbirth. However, the things that the pregnant Jenny_t1 may prioritise and value could be significantly different from those of her later self. That is because some life events can profoundly alter our perspective. Such events are sometimes described as ‘transformative’. The philosopher Laurie Ann Paul provocatively imagined becoming a vampire. (Paul, P1–4)16 Overnight, someone’s way of life, their viewpoint, their values and preferences would transform. Such an experience might radically undermine our ability to make informed choices (because of our difficulty in imagining what life would be like). But on the Reductionist Identity account, it might also suddenly and significantly weaken the psychological connections that are morally significant. ‘[W]hen there has been a significant change of character, or style of life, or of beliefs and ideals—we might say, ‘It was not I who did that, but an earlier self’. (Parfit, p305)4. Paul cites becoming a parent as a paradigmatic example of a transformative experience.17 In Jenny’s case there might be the additional, even more profoundly transformative, experience of bereavement. A news article describing a real case of a free-birth choice that ended badly, cites a woman whose baby died following planned free-birth. The mother described vividly her subsequent guilt, and her conclusion in retrospect: ‘I think I brainwashed myself with the internet’.18 Her description of a profound shift in perspective marries with the notion that the self who experiences the harm in such cases might be different, to an important degree, from the earlier self who made a harmful choice.¹⁶

On an Identity Relative Paternalistic account, the reasons to be paternalistic would be stronger for Advance resuscitation refusal and Hospital birth refusal, than for Vaccine refusal. Since we would not permit harm to other people in the third-party variations of those cases, we potentially should not permit John and Jenny to refuse treatment. Or, at the very least, we should be more inclined to overrule or disallow their decisions. However, that would not apply to James’ Vaccine refusal.

In contrast, a time-neutral paternalistic account would treat these cases as symmetrical. Since we would not allow third parties to harm other people in these ways and for these reasons, we should allow neither James nor Jenny nor John to harm themselves in the ways that they intend.

If we take a reductionist approach to identity, which version should we adopt, which is most plausible?

There is some reason to think that Parfit would have supported Identity relative Paternalism. The reductionist identity morality and paternalism claims as articulated by Parfit relate to the harm that individuals do to their ‘future selves’. (Parfit, p320–1)4 In support of these claims he provided examples where the prudential unity relations are weakened. For example a boy starting to smoke and causing great suffering fifty years later. (Parfit, p319)4 However, it is worth being clear whether and why the moral reasons not to harm oneself apply only to far future selves and not our more proximate selves.

Parfit gave two alternative ways of expanding the scope of moral theory to include harm to our future selves. The first would be impersonal. The moral reason not to self-harm is because it results in reduced overall well-being or greater suffering. From an agent-neutral consequentialist perspective, the loss of well-being due to self-harm is the same as the loss of well-being incurred when a third party is harmed.¹⁷ The second alternative is agent relative. We could expand our understanding of our special duties and obligations (to kin, to friends, to clients) to include a duty to the self.¹⁸

If we expand the scope of morality in these ways, that would provide a basis for the reductionist identity moral claim and the corresponding paternalism claim. Yet on either basis, that would potentially apply equally to the proximate future self and the distant future self. From an impartial consequentialist perspective it would be just as harmful for John to make an unwise advance decision that shortens his life if the harm accrues shortly after his advance directive or many years later.¹⁹ Likewise, it is not clear why our agent relative duty to self applies to our far future and not our near future self.

If the moral reasons not to harm do not change over time, or with weakening of psychological connections, that would appear to support Time-neutral Paternalism. But it would be worthwhile returning to why the Reductionist Identity account provides support for paternalism in the first place.

When we recognise that what matters to us (in terms of future selves) is a matter of degree, that may change what we have reason to care about in an egoistic way. We may then come to care less about harms that occur to ourselves in the far future. It would not be irrational to make imprudent decisions if the harms will occur at a time later when the psychological connections to our current self will be relatively weak. (Parfit, p313–4)4 The Reductionist Identity account weakens the prudential reasons that we have to avoid harms to future selves. But it does not, itself, generate a corresponding moral reason to intervene.

Where does the case for paternalism come in then? As argued, the moral reasons not to harm, are not identity relative, they are time-neutral. What the Reductionist Identity account does is not to create a moral reason to avoid harm to a future self—rather it potentially unmasks those moral reasons. Figure 1 illustrates this.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

The strength of reasons to avoid harm and the time point of harm occurring. Prudential reasons are “reason(s) from the prudential point of view”.28 The moral reasons to avoid harm do not change over time, but the prudential reasons may diminish (where prudential unity relations are weaker). The individual does not necessarily regard their actions as harmful, and may regard it as in their self-interest to eg avoid a transfusion or give birth away from a hospital. However, those self-interested (prudential) reasons potentially attenuate in strength the further in the future the self that would be impacted by the decision.

As the figure indicates, for proximate self-harming, there are both prudential and moral reasons not to harm the near-future self. For far-future self-harming, the prudential reasons (currently) may be relatively weaker insofar as the psychological connections and continuity are diminished, while the moral reasons remain the same.

On this model, it seems that the reductionist identity moral claim should be time neutral. We ought not to do to our future selves (whether or not we are closely psychologically connected to those future selves) what it would be wrong to do to other people. But there is a further question about the Reductionist Identity Paternalism claim. That is because for proximate future selves there is a potential conflict between the prudential reasons not to harm (as perceived by the individual) and the moral reasons. We need to consider then the difficult question of how to balance prudential and moral considerations.

Paternalism and the duality of practical reason

Sometimes we face choices between what we believe would be best for us, and what would be impartially best. For example, it may be that we could help someone else, but only at a significant personal cost. Or it may be that we could take a course of action that would be better for ourselves, but at the cost of failing to do what we morally (impartially) ought to do.

The moral philosopher Sidgwick regarded the potential conflict between these two different types of reasons as ‘the profoundest problem of ethics’. (Sidgwick, p386)20 The problem, as Sidgwick saw it, was that such reasons are not straightforwardly comparable. It is not clear how much sacrifice of personal well-being we are required to make for the sake of impartial beneficence. There is no external viewpoint that would allow us to answer such a question. (Parfit, p133–4)21

According to Sidgwick, if we have a choice between what would be impartially best and what would be best for ourselves, it would be rational to take either choice—this is what he referred to as the ‘dualism of practical reason’.²⁰

The dualism might not be thought to apply to paternalism cases like the first-party treatment refusal cases cited above. In these cases, the prudential reasons to avoid harm and the moral reasons actually coincide. Indeed, the reason that we might be tempted to be paternalistic is because of concern for the individual’s well-being. So there is no conflict between incompatible reasons.

However, from the point of view of James and Jenny and John, the prudential reasons do diverge. They believe that the decisions they are making would be best for their current and future selves. They believe that they would be harmed by accepting the vaccine, transfusion or hospital birth. They appear to be mistaken, but from their perspective that is not the case. (Indeed, they are not likely to even recognise this conflict since they would regard the choices they are making as promoting the well-being of their future selves).

The question for paternalism is whether health professionals or societies are justified in overruling an individual’s personal judgement about what would be best for himself on the basis of a reliable concern that this would in fact cause the future person great harm. This does seem to potentially be a conflict between a type of prudential reason and a moral one. (A familiar way in medical ethics of characterising this conflict would be to see this as a clash between ‘autonomy’ and ‘beneficence’).

One response, drawing on the dualism of practical reason, would be to say that there is no way to weigh up or arbitrate between these two types of reason. Sidgwick, himself, could identify no way of balancing these reasons. In that case, faced with any degree of conflict between prudential and moral, autonomy and beneficence concerns, we may decide to give primacy to autonomy and reject paternalism. Indeed, the conventional response to treatment refusal cases takes exactly that path. No matter how great the well-being cost, or how irrational the reason, so long as the individual has the capacity to decide, they should be permitted to refuse treatment.

Yet, many philosophers who have followed Sidgwick have rejected his pessimistic conclusion that no answer could be found that would balance the two types of reason. Parfit’s own discussion of Sidgwick emphasised that he thought that Sidgwick’s account depended on the rational significance of personal identity.

Given the unity of each person’s life, we each have strong reasons, Sidgwick claims, to care about our own well-being in our life as a whole. And given the depth of the distinction between different people…one person’s loss of happiness cannot be compensated by gains to the happiness of others.[32, 133]

But this was, according to Parfit, to overstate the importance of personal identity. (Parfit, p136)21 On the reductionist identity account, our reasons to care about our future are based not on the fact that such a future is ‘ours’ - rather on the basis of the psychological relations between our current and future selves. Moreover, Parfit contended that we can have some similar partial reasons to care for the well-being of other individuals connected to us (friends and relations) and separate impartial reasons to care about everyone’s well-being.

Parfit’s own preferred way of addressing the dualism was through what he called the wide value-based objective view and the notion that such reasons are comparable, although only imprecisely.

Wide value-based objective (WVO) view If one of two possible acts would make things go impartially better, while the other act would make things go better from a partial perspective (for ourselves or someone close to us), we could have sufficient reason to act in either way. (Parfit, p137)21

Drawing on the WVO view, Parfit concluded that strong impartial considerations could sometimes outweigh weak prudential ones. We would have much stronger reasons to save many strangers from death or agony than to save ourselves from some minor harm. (Parfit, p137)21 For a practical example, on one plausible interpretation of this view, a passing stranger would have stronger reason to save a child drowning in a pond, than to save his expensive suit.²¹

Parfit alluded to the relevance of a Reductionist Identity view for the dualism and as support for the WVO view. However, it also seems that we could draw on this WVO view in thinking about paternalism. The WVO is about moral as well as rational permissibility. One problem with time-neutral paternalism is that it appears to give no weight to the prudential reasons for acting—it takes only an impartial perspective. In contrast, identity relative paternalism would defer to the individual’s judgement for harms that are proximate, but potentially give greater weight to impartial moral considerations as the psychological connections between current and future-self diminish. This would potentially be supported by the idea, on the WVO, that strong impartial reasons could outweigh weaker prudential reasons. That suggests that paternalism would be most justified in cases where the harm will be accrued by a future self relatively psychologically distant from the current person.

Practical implications of identity relative paternalism

I have defended Identity relative paternalism. What would such a view mean for the harm principle and for refusal of medical treatment?

According to the view I have described, if we would not allow an individual to refuse treatment for a third party (eg, where they are a surrogate decision maker), we should potentially disallow refusal of treatment where the harm will be accrued by a future self. How strong a reason there is to act paternalistically will depend on the relative strength or weakness of prudential unity relations. The sharp boundary between harms to self and harms to third parties would be dissolved.

Objections