IAIN LAW
AUTONOMY, SANITY AND MORAL THEORY
ABSTRACT. The concept of autonomy plays at least two roles in moral theory. First, it provides a source of constraints upon action: because I am autonomous you may not interfere with me, even for my own good. Second, it provides a foundation for moral theory: human autonomy has been thought by some to produce moral principles of a more general kind. This paper seeks to understand what autonomy is, and whether the autonomy of which we are capable is able to serve these roles. We would naturally hope for a concept of autonomy that is value-neutral rather than value-laden. That is to say, we would want the judgement that I am autonomous to depend wholly on, say, structural features of my psychology, and in no way to require us to make ethical judgements, or other value judgements. Being value-neutral is perhaps desirable in a concept of autonomy serving the first role, and plausibly indispensible in one playing the second. I shall argue, however, that value-neutral conceptions of autonomy are impoverished and out of line with our intuitions; set out and defend an explicitly value-laden conception of autonomy; and explore the implications of such a view for the ability of autonomy to play the roles mentioned above. KEY WORDS: autonomy, constraint, freedom, morality, sanity
I Many of us suppose that it is permissible to do things to them which children and the mentally infirm do not want, so long as these things are done for their own good. For example, we think it permissible to administer medicines to children without reference to their wishes. That someone with an eating disorder refuses treatment does not necessarily settle the question of whether he or she will be given it. We tend also to suppose that taking such action is impermissible in the case of mentally fit adults: if I refuse some treatment, my doctors would be wrong to administer it anyway, even if it would improve my health. A prominent recent case of this kind was that of ‘Miss B’, a paralysed woman who wished her life-support system to be removed. The High Court judged that this wish was one that her doctors were bound to respect.1 Why do we make such judgements?
1 This case was widely reported in the British press. See for example The Guardian,
23rd March 2002, 1. Res Publica 9: 39–56, 2003. © 2003 Kluwer Academic Publishers. Printed in the Netherlands.
40
IAIN LAW
The answer some implications of which I shall explore in this paper is that mentally fit adults are autonomous, while children and (at least some) mentally ill people are not. Our possession of autonomy is thought by many to impose constraints on what may permissibly be done to us. Even more, it is thought by some that human autonomy can be used as one of the foundations of a moral theory; that the fact that we are autonomous has implications which can result in a more or less complete system of principles governing what we can and cannot do to one another (for example, Julia Annas makes this sort of thought central in her criticisms of Plato2 ). Understanding what autonomy is, who possesses it and how it is related to moral values and constraints is thus a central issue. I shall argue that the best available conception of autonomy will not serve the second of these roles, and serves the first in a different way than is commonly supposed. Starting by showing that value-neutral conceptions of autonomy are flawed, I shall go on to set out and defend an alternative, value-laden conception of autonomy; and then discuss how that conception works in theory and in application. One of the most prominent advocates of the view that autonomy is a source of constraints on our actions is Gerald Dworkin, advocating a view of autonomy known as the hierarchical model.3 This hierarchical model of autonomy is attractive in that it seems to capture much of what we intuitively think autonomy is; namely a freedom of a particular kind, thought to be similar to such things as self-control or self-possession. One way to look at it might be to say that autonomy is freedom not only to act as we will, but to choose the ends for the sake of which we act. As such, autonomy can be thought of as the ability to scrutinise the things for the sake of which we act – our projects, goals and desires – and decide whether or not they are ends we really want to have. The hierarchical model of autonomy expresses this way of thinking in terms of orders of desire. A first-order desire is one which has some state of affairs in the world as its object: a desire that you have ice-cream, for instance, or a desire that someone trip and fall. Second-order desires have first-order desires as their objects: a desire not to want ice-cream, for example, or not to want such horrible things to happen. According to the hierarchical model, autonomy 2 Julia Annas, An Introduction to Plato’s Republic (Oxford: Clarendon Press, 1981), 172–8. 3 Gerald Dworkin, ‘Autonomy and Behavior Control’, Hastings Center Report 6 (1976), 23–8; ‘The Concept of Autonomy’, in ed. John Christman, The Inner Citadel (Oxford: Oxford University Press, 1989) 54–62; and The Theory and Practice of Autonomy (Cambridge: Cambridge University Press, 1988). See also the similar model set out in Harry Frankfurt, ‘Freedom of the Will and the Concept of a Person’, Journal of Philosophy 68 (1971), 5–20.
AUTONOMY, SANITY AND MORAL THEORY
41
is, inter alia, a matter of there being harmony between those desires that you act upon (your volitions) and your second-order desires. Dworkin calls such harmony ‘authenticity’.4 In other words, it is a necessary condition of autonomy that your volitions be ones that you want to have. The simplest sort of hierarchical model would be one which regarded authenticity as all that is required for autonomy. Such a ‘basic hierarchical model’, however, is not feasible: authenticity is not a sufficient condition of autonomy, since it is possible that an agent endorses all her first-order volitions, but nevertheless is not autonomous. Some ways in which people acquire the standards by which they assess their first-order desires are such as to render them non-autonomous: they may, for example, have been brainwashed, or hypnotised or systematically deceived in certain ways. Imagine someone who endorses the volitions she has to buy chocolate, but who does so only because her second-order desires have been conditioned by manipulative chocolate salesmen. This person has authenticity – there is no conflict between what she wants and what she wants to want. She is, however, not the kind of person whom we would think of as autonomous (at least in her chocolate-buying behaviour). To overcome problems of this sort, Dworkin appeals to the notion of ‘procedural independence’. If a person has authentic desires, and came by them in a procedurally independent fashion, then they are autonomous: Spelling out the conditions of procedural independence involves distinguishing those ways of influencing people’s reflective and critical faculties which subvert them from those which promote and improve them. It involves distinguishing those influences such as hypnotic suggestion, manipulation, coercive persuasion, subliminal influence and so forth, and doing so in a non ad hoc fashion.5
This seems an excellent description of what an account of procedural independence would have to achieve. Unhappily, however, no such account has appeared, for it seems difficult to give any content to the notion without starting a vicious infinite regress. It would be natural to say that the secondorder desires have been acquired in a procedurally independent fashion if they were chosen freely by the agent in question. What would it mean to say that they were freely chosen, however, except precisely that their coming to be the agent’s desires was endorsed by the agent? This would just force us to ask the question again of the standards by which the agent made that choice, and so on. Thus, the most natural thought about what makes a desire one which was acquired independently is one which is fatal to the hierarchical model. But no other criterion of procedural independence presents itself. So Dworkin’s attempt to give necessary and sufficient 4 Dworkin, ‘The Concept of Autonomy’, op. cit., 61. 5 Dworkin, The Theory and Practice of Autonomy, op. cit., 18.
42
IAIN LAW
conditions of autonomy that stipulate only how desires should be acquired, and not what those desires should be, fails.6
II But the basic hierarchical model can be employed as a component of a more promising account. Dworkin’s model sees autonomy as a matter of my desires being under the control of my deeper or truer self. For this reason, Susan Wolf calls Dworkin’s account, and others like it, the Deep Self View.7 Her own view she calls the Sane Deep Self View,8 since she believes that what needs to be added to authenticity is not procedural independence but sanity.9 In other words, it is not enough that your desires be controlled by your deep self: that deep self must also be sane. What does this mean exactly? Wolf illustrates her view by describing a cruel and vicious man, Jojo, who endorses all his volitions, but does so because his upbringing has left him incapable of doing otherwise. According to the basic hierarchical view, Jojo is autonomous; but, Wolf claims, ‘it is dubious at best that he should be regarded as responsible for what he does’.10 One reason for doubts on this score is that it seems likely that anyone else brought up in the fashion in which he was raised would have become the same sort of vicious and evil man. But, according to Wolf, this reason for saying that Jojo is not responsible would lead to the conclusion that no one can ever be responsible, since ‘in the last analysis, it is not up to any of us to have the deepest selves that we do’.11 However, there is another way of explaining why Jojo is not responsible which will preserve the possibility of responsibility for the rest of us. ‘Not all the things necessary for freedom and responsibility’, Wolf says, ‘must be types of power and control. We may need simply to be a certain way, even though it is not within our power to determine whether we are that way or not.’12 6 For a more detailed discussion of this, see Iain Law, ‘The Hierarchical Model of Autonomy’, Cogito 12 (1998), 51–7. 7 Susan Wolf, ‘Sanity and the Metaphysics of Responsibility’, in ed. John Christman, op. cit., 137–51, p. 140. 8 In later work she calls it the Reason View: Susan Wolf, Freedom Within Reason (Oxford: Oxford University Press, 1990), passim. 9 Wolf, ‘Sanity and the Metaphysics of Responsibility’, op. cit., 145. 10 Ibid., 143. 11 Ibid., 144. 12 Ibid. It should be noted that Wolf’s is an account of moral responsibility, which is not my concern here. I wish to borrow elements of her account in order to construct a ‘Wolfian’ account of autonomy, understood as the sort of concept discussed so far.
AUTONOMY, SANITY AND MORAL THEORY
43
The way we need to be, according to Wolf, is sane. She means this in a quite specific sense, defining sanity as having two constituents. The first is that one’s beliefs be accurate, i.e. represent the real world to one as it is. The second is that one’s values similarly give one an accurate picture of what the world is like. The definition she gives of this notion of sanity is that it is ‘the minimally sufficient ability to cognitively and normatively recognise the world for what it is’.13 This account will, she claims, enable us to distinguish between Jojo and ourselves as far as our moral responsibility goes, which we could not have done on the simple deep self view, or on any view that requires that the deep self be under the control of some deeper self, ad infinitum. I propose to adapt Wolf’s Sane Deep Self View as an account of autonomy. What a Wolfian account of autonomy will require, then, is that one’s first-order desires be scrutinised and checked by one’s higher-order desires and values (one’s ‘deep self’) and also that these higher-order desires and values be ones held by a person who is in Wolf’s sense sane. Sanity is necessary for autonomy. Alone, however, it is insufficient, since the absence of authenticity would mean that although your deep self was sane, your first order desires were not under the control of this deep self. You would know that certain activities were valueless (or worse), but be unable to control your desires to indulge in them. Sanity needs authenticity to become autonomy: jointly they are its sufficient condition. But is it possible to give an account of autonomy which rests upon the condition that to be autonomous one must also be sane? Two important considerations that seem inimical to such a possibility present themselves. First of all there will be a problem in formulating the condition of sanity more exactly. Wolf’s description of it as ‘the minimally sufficient ability to cognitively and normatively recognise and appreciate the world for what it is’ means that to give any real substance to the notion of sanity, one would have to give some specification of what minimum level of cognitive and normative truth about the world the (barely) sane person must reach. But no such specification could ever be uncontroversial, not only because of disagreement about how far into cognitive or normative error a person could fall without thereby coming to be counted as insane, but also, and perhaps more crucially, because of deep divisions and disagreements about what the truth is, whether cognitive or normative, about the world. Furthermore, even if this difficulty could be overcome, there is a problem in that autonomy and sanity at least appear to be conditions that pull against one another. Autonomy is something to do with the ability to choose one’s own ends freely, but sanity involves being constrained by the world or by the 13 Ibid., 145.
44
IAIN LAW
truth in the range of ends one can adopt. Is there any way of reconciling these seemingly opposite conditions? I shall examine these two problems in turn.
III If the concept of autonomy is to be useful to us when we consider, for instance, whether or not a person should be deprived of the right to run her own life, then it seems obvious that we shall need to be able to tell whether or not that person is autonomous. If sanity is part of what it is to be autonomous, then we shall also need to be able to determine whether or not she is sane; and to do this we shall need a workable criterion. Even if we were confident about the prospect of devising such a criterion insofar as sanity involves having true non-normative beliefs (and that would be hard enough) this will lead us into difficulties when it comes to that part of sanity involving, in Wolf’s words, ‘appreciating the world normatively for what it is’: in order to find someone sane, we would have to look at their scheme of values and judge that they had the right ones. A major problem here is the obvious charge that we are just assuming that our own values are the correct ones, and branding anyone who disagrees with us as insane. So we need some way of identifying what the correct values are if we are to employ Wolf’s sane deep self account – and that, to say the least, is a bit of a tall order.14 Perhaps the situation is not as bad as it first appears, however. Wolf’s definition of sanity was of a ‘minimally sufficient ability’ to recognise the true values. Given that we are looking for a minimum condition, we may not have to be able to give a full account of the values that are correct, but just be able to identify when someone falls so far short of having the ability to recognise them in their evaluative beliefs and attitudes that they are insane. This is the approach that Wolf herself adopts. Rather than give an account of sanity and demonstrate that Jojo is not sane, she simply 14 It might seem that there is another interpretation of the Sane Deep Self view which
avoids this problem. One might think that what values a person has are irrelevant to determining whether or not they are autonomous. Rather, what matters is whether their desires reflect whatever values they in fact have. This sort of interpretation, however, will face the very problems it was hoped that the Sane Deep Self view would avoid. For we can imagine a person who has harmony between their values and desires, but who came to have their values in such a way that we would not think of them as autonomous. Obviously, this is the same sort of problem which led Dworkin to add procedural independence as a condition of autonomy, and has led me to look for a better account of autonomy, and one which incorporates sanity as a necessary condition.
AUTONOMY, SANITY AND MORAL THEORY
45
says of him that ‘a person who, even on reflection, can’t see that having someone tortured because he failed to salute you is wrong plainly lacks the requisite ability’.15 This sort of approach may serve well enough if we are prepared to say of something that we don’t know what it is but that we know it when we see it, but in this case I think the question is whether we can provide any justification for our certainty that we are right and that they are wrong. If we are reduced to saying that their behaviour just shows them to be insane, and that is all there is to it, then the charge of just assuming and imposing our own values remains. We could spend a lot more time exploring what might be involved in having this ‘minimally sufficient ability’, but whatever conclusions we might reach, there will always remain this problem: it is simply a matter of fact that there are many people to whom it is just not obvious that there are even basic moral truths in the way that there are non-moral truths. Faced with a person who saw nothing wrong in torture, such “moral nihilists” would say that the person was not making any sort of mistake (although they might be quite happy to express disapproval of such a person). It would appear, then, that the Sane Deep Self View has to simply insist that it is right by fiat or give up: interestingly, however, Wolf herself does not appear to be quite so pessimistic. Wolf argues that her view is not in fact committed to any particularly strong meta-ethical claims. Her argument takes the form of a demonstration that nearly every kind of meta-ethical position can accommodate her view of sanity without difficulty.16 She points out that just because the idea of sanity involves the idea of being able to recognise the values that there are, that does not mean that one must be committed to thinking that values are universal. Anyone who holds any kind of substantive moral view whatever will be able to accommodate the idea of sanity, since they will just see sanity as recognition of whatever they themselves take to be the normative case. This extends even to the view which Wolf calls nihilism: the belief that reason is neutral about questions of value and that therefore to think that any situation, action or state of affairs is any better than any other is a mistake. As Wolf sees it, this is itself to make a substantive evaluative claim, namely that one ought to be neutral on all questions of value; and hence the nihilist will be able to accept the idea of sanity, since as far as they are concerned, a sane person will be one who is able to apprehend the truth about value, namely that there is no reason to value anything over anything else. Wolf contrasts such nihilism with what she calls ‘conceptual subjectivism’, which she says is the only meta-ethical 15 Ibid., 145. 16 Wolf, Freedom Within Reason, op. cit., 118–29.
46
IAIN LAW
position which is not able to accommodate the idea of sanity. Conceptual subjectivism rejects the conclusion drawn by the nihilist that there is, objectively, nothing to choose between ends: nihilism gets it wrong, according to the conceptual subjectivist, by even trying to make objective value judgements such as the value judgement that there is nothing to choose between kindness and cruelty: According to the subjectivist, Reason is silent, not neutral, on questions of value. Not only does it fail to support one value judgement over another; it fails to support one value judgement over another, including the judgement that anything goes. Thus where the nihilist would say that ‘killing is bad’ is no better justified than ‘killing is good’, the subjectivist would add that ‘killing is neither bad nor good’ is no better justified than the other alternatives.17
Conceptual subjectivism holds that it is simply inappropriate to make judgements about value at all. Indeed, it seems that conceptual subjectivists are committed to supposing that to talk of making judgements about value is to attempt a logical impossibility. Making a value judgement, where doing so implies the ascription of a truth value to the content of the judgement, is just a category mistake. One may value things in the sense of having them as one’s ends, but to try to make value judgements is simply wrong-headed. Thus we can see that the conceptual subjectivist, alone of all meta-ethical theorists, will be able to make no use at all of Wolf’s idea of sanity, for the very notion of normative competence crucial to the idea of sanity involves the same category mistake as trying to make a value judgement. It requires that moral opinions and attitudes be truth-apt; but the conceptual subjectivist believes that they are not. It is perhaps slightly disingenuous of Wolf to say that only one metaethical theory is incompatible with her Sane Deep Self View. For what she refers to as ‘conceptual subjectivism’ would seem to be a blanket term covering several of the last century’s popular meta-ethical theories.18 Having said that, her point is well made that many more meta-ethical positions could be made compatible with the Sane Deep Self View, with its reliance on the notion of normative competence, than might have been thought. Certainly, and quite surprisingly, a position involving the belief that no action, intention, end or state of affairs is any better than any other is perfectly compatible with acceptance of the Sane Deep Self View. Unfortunately for our purposes, this attempt by Wolf to make her view as widely acceptable as possible serves only to reveal more clearly the difficulties in making it workable. The Sane Deep Self View requires that we 17 Ibid., 129. 18 Such as emotivism, prescriptivism and the like.
AUTONOMY, SANITY AND MORAL THEORY
47
accept the notion of normative competence. To be normatively competent one must be equipped to discern and be guided by the moral facts, whatever they may be. Wolf’s worry was that this would be incorrectly taken to involve a commitment to a particularly strong form of objectivism regarding substantive moral claims about the world. In order to counter this danger, she argues that her view should be acceptable to any theory which accepts the truth of evaluative statements of any kind, including even theories which hold it to be true that nothing is any better or any worse than anything else. But the worry with which Wolf attempts to deal here is not the same as the one with which we have been struggling previously. That worry involved determining who is autonomous and who is not. As a result, it involved the need to be able to determine the sanity and normative competence of individuals. The problem is that in order to be able to determine who is sane and who is not, we shall need to know which of the various meta-ethical and normative theories compatible with the Reason View is actually correct, for when we are judging whether a person’s views are too far from the normative truth, we shall need to know what that truth is ourselves.
IV Before considering the implications of this for the concept of autonomy and its uses in justifying constraints on action and serving as a foundation for moral theories, however, I would like to look at the second question: whether introducing a condition of sanity as one of the necessary conditions for being autonomous is antithetical to the very spirit of the idea of autonomy. I want to get this out of the way at this point, since if it were to turn out that autonomy cannot be the kind of thing I am arguing for, the issues I want to go on to would be rendered moot. To know whether those issues are indeed ones we need to deal with, we need first to know whether my preferred conception of autonomy is defensible as such. John Christman has argued that it is not: But adding objective or ‘external’ conditions of rationality to the requirements for autonomy (conditions regarding optimal evidence gathering, for example) may in fact conflict with the concept of self-government that autonomy is intended to express. (This is not to mention the indeterminacy of a demand like Wolf’s – that an agent have the correct moral view – in light of the tremendous lack of general agreement over even fundamental questions of morals and value.) On the view that only rational, fully informed selves are autonomous, it follows that the most fierce and uncompromising interferences with a person’s value-judgements, desire formation, or thought patterns are not interferences with autonomy at all if those values, desires, or thoughts are irrational ones. As long as the interfering agent has beliefs that are better supported by the objective evidence, then no loss
48
IAIN LAW
of autonomy is suffered from any form of manipulation for the sake of more reasonable beliefs and desires.19
Christman is voicing a broader objection than one just to Wolf’s particular view. He is arguing that any sort of ‘external’ condition on the requirements for autonomy must be misplaced, whether it be a condition that the agent must be sane, in Wolf’s sense, or whether it be that they display various kinds of rationality. Any specification which strays from the path of laying down only conditions that are internal to the agent – such as Dworkin’s procedural conditions – must be going wrong. I can discern two separate arguments for this conclusion. First, there is an apparent belief that if you start including external requirements in the necessary conditions of autonomy, then you have ceased to be specifying conditions of autonomy, but have strayed into discussing something else. The thought is that the very idea of autonomy contains the notion that there are no external conditions on its possession. This objection has some force. As Christman points out, the idea of autonomy is the idea of some kind of self-government. We might well think that one of the important rights associated with the possession of autonomy is the right to get things wrong, to make mistakes, but to make them for ourselves. This is an ideal which external conditions appear to threaten. Wolf herself is quite explicit about this: being sane, she says, requires that ‘one’s self be controlled by the world in certain ways’.20 So one way in which one might want to object to external conditions, and hence to a Wolfian account of autonomy, would be to say that if your conception of autonomy involves external conditions, then it has ceased to be a conception of autonomy, whatever else it may be. Second, there is the worry that a conception of autonomy which contained external conditions would be a morally bad conception in practice. If it is legitimate to interfere with non-autonomous people for their own good, and displaying ignorance or hostility towards what your own good is makes you non-autonomous, then the Sane Deep Self view, and other ‘external condition’ accounts like it, could be used to justify radical interference with vast numbers of people. I shall consider briefly both of these objections. There are two ways in which someone defending an account of autonomy which included external conditions could respond to the charge that they have gone beyond the bounds of the concept of autonomy. They could deny that this is the case, citing uses of the concept that might be more consistent with their account than with accounts lacking external conditions. Alternatively, they could admit that what they were doing 19 John Christman, ‘Introduction’, in Christman, ed., op. cit., 3–23, p. 12. 20 Wolf, ‘Sanity and the Metaphysics of Responsibility’, op. cit., 144.
AUTONOMY, SANITY AND MORAL THEORY
49
might stretch the concept of autonomy beyond its traditional limits, but argue that such a step was necessary because of shortcomings in that traditional conception. I believe that both of these responses are open to me. First, it seems that accounts of autonomy which do not contain external conditions leave us having to admit that people we naturally regard as not being in control of their lives are in fact autonomous. Recall that the only account of autonomy left open other than one with ‘external’ conditions is what I earlier called the basic hierarchical view. Dworkin’s attempt to supplement that model with the further internal condition of procedural independence, I argued, is an empty gesture. This means that we are faced with a class of people whom we would intuitively judge to be non-autonomous, but lack any internal criterion by which to make that judgement. This class will include people like the manipulated chocolatelover mentioned earlier, as well as people who have apparently lunatic desires but who are entirely lacking in conflict or internal error or strife with respect to those desires. Faced with someone who wishes, say, to devote all their time to counting the grains of sand on Luskentyre, and who exhibits all the characteristic signs of having considered this goal and endorsed it fully, what are we to think? If we deny ourselves an external condition, the options are these. Option one: the man is autonomous. Whether this option seems attractive is, I suggest, largely a function of how willing we are to think there is nothing in any way objectionable about having such a desire. Option two: despite appearances, what he wishes to do reveals a lack of autonomy. In some hidden way, he does not really want to count grains of sand. Option two, it seems to me, is a cop-out. It is a covert appeal to an external criterion, namely the unreasonability of the man’s desires. Unwilling to acknowledge that this is what we are doing, we assume that the bizarre desire is evidence of some internal flaw – but this is an unwarranted assumption. In real life cases, we may rightly worry that such dubious reasoning is used to justify administering treatment to people who have odd desires but who are otherwise apparently rational: people with serious eating disorders come to mind. If we wish to conclude that such people lack autonomy and may therefore be treated against their wishes, the basic hierarchical account will not help us. Some will perhaps object that there is an alternative account that may help us. They will see the issue of autonomy as being one with the issue of rationality – if (and only if) someone is rational, they are autonomous. But the same issues arise for any such account, since rationality too can be understood either as having only internal conditions or as also having external ones. The advocate of this approach cannot be supposing that
50
IAIN LAW
rationality involves some sort of apprehension of the good or right, for then there would be no useful distinction from the view under discussion. Any view according to which being immoral means being irrational is ruled out. Wolf seems sometimes to share such a position: the alternative name she gives her position, ‘the Reason View’,21 itself implies a close connection between rationality and sanity. A link of this kind is one which a putative rival ‘rationality view’ must reject, or it will share the value-laden nature it is intended to avoid, and thus not be a genuine alternative. The point is that the rationality in question must be a kind of rationality that even someone who was not sane (in Wolf’s sense) could possess. So the idea of rationality at issue is an internal one, involving consistency, choosing appropriate means to given ends, and so on. It is far from obvious, though, that this can be made to work. Can rationality and sanity be so easily severed? Being rational involves what one might call ‘plasticity’ – the capacity to revise one’s beliefs in the light of relevant evidence. Someone who happens by chance to have no false beliefs, but who lacks the ability to consider whether his beliefs might require revision cannot be rational. Plasticity must be distinguished from mere malleability, though – someone who is too willing to change his mind may also exhibit irrationality. Making this distinction requires us in turn to distinguish between good and bad reasons for altering one’s beliefs. Now think about how a rational person treats her values. Will not a similar degree of plasticity with respect to one’s normative beliefs be a condition on any reasonable account of rationality – any account, at least, that is plausibly so closely involved with autonomy?22 Suppose that you encounter someone who manifests plasticity with respect to her factual beliefs, but who wholly lacks it with respect to her values. What does the rationality view say? If it says that she is irrational, then the rationality view shares the meta-ethical commitments of the Sane Deep Self view, since it would be supposing that there are good and bad reasons for changing one’s values. So to be interestingly distinct, the rationality view must accept that the person is rational. But this is implausible. It involves not only an uncomfortably sharp distinction between facts and values, but also an etiolated view of rationality itself. The Sane Deep Self view therefore seems likely to accord far more closely with the extension of the concept of autonomy as it is actually used than that extension accords with value-neutral, non-external condition 21 Wolf, Freedom Within Reasons, op. cit., passim. 22 I make this qualification only because I suppose it is possible that someone might
want to construct an account of something they called ‘rationality’ which took no account of normative beliefs at all. While this may be possible, any such faculty would not be a plausible candidate for the role sketched here.
AUTONOMY, SANITY AND MORAL THEORY
51
accounts such as Dworkin’s. But it could still be argued against my view that it cannot be an account of autonomy since the concept of autonomy involves governing oneself, and my account involves being governed by something else, namely the true and the good. In response to this I can simply admit that this is true, but insist that my account makes more sense and works better than accounts which try to make sense of autonomy in terms of pure self-government, and which are open to the infinite regress attack. If my account of autonomy must be admitted to be a substantial revision of the more commonly accepted concept of autonomy, it can at least be argued to be a necessary one. I can see no reason to reject the account of autonomy I have proposed because it threatens personal liberty. Accounts of autonomy are intended to provide a criterion by which we can separate those people with whom it is wrong to interfere even for their own good from those with whom it is permissible so to interfere, should their good require it. There is nothing in the account given to make it inevitable that it could be used to justify greater interference than other accounts. Whether or not it does so will depend not on the Sane Deep Self view but on the system of values with which it is coupled, and the attitudes of those who implement it. Indeed, I wonder whether if we stopped treating what are in fact value-judgements as value-neutral judgements about autonomy, we might not come to be more reluctant to act against the will of the non-autonomous. If I perceive the anorexic as having no internal, procedural flaw in her desires and beliefs but rather as wanting and valuing different things from me, will I not be more likely to hesitate to impose treatment on her, for all that I think she is not acting autonomously? This brings me to the issue of application.
V Suppose that we have to decide whether or not to administer some medical treatment to a person who does not want to have it. We need to determine whether or not this person is autonomous. We therefore need to know (a) whether her first-order desires are under the control of her ‘deep self’ and (b) whether her deep self is sane, where sanity involves, among other things, having (some) correct moral views.23 The position seems to be this. We would have to know which moral views are correct, and have some view as to how many erroneous moral beliefs one may have before one reveals oneself to be insane. Armed 23 To what extent we require that her moral views be correct, i.e. how much error we
allow, is discussed below.
52
IAIN LAW
with these standards, we could carry on using autonomy as a criterion for deciding whether or not to respect patients’ wishes. But using autonomy as such a criterion is very likely now to appear to be a more dubious procedure, and I think this appearance is correct. One would have to be very confident in one’s opinions about (a) and (b) above before one could feel entirely comfortable in making such decisions. I shall take the two issues in reverse order. Just now I wrote of having the correct moral beliefs as a criterion of autonomy. But sanity, as defined by Wolf, involved not having correct beliefs, but rather having the capacity to form correct beliefs. I think it is justifiable, however, to employ the former condition in judging whether the latter is present. Some capacities may be present in us un-manifested, but the capacity to see the world correctly in normative terms is, I suggest, not among these. In any remotely normal human life there will be countless occasions upon which such a capacity would be called upon. Failing to have any correct moral beliefs is therefore always evidence of failing to have the capacity to form correct moral beliefs, rather than evidence of that capacity’s not having been actualised. The next question is what proportion of false moral beliefs is necessary for us to draw the conclusion that someone is insane. It seems that no strict proportion will be available: the presence of even one erroneous belief, if it were of sufficient importance, might be enough to convince us of this, as Wolf’s own example of Jojo shows.24 We might also recall Anscombe’s famous claim that anyone who thinks that it might be right to execute an innocent man ‘shows a corrupt mind’.25 (Of course, while one serious error might be enough to convince us the person was insane, we would not be likely to suppose that this was the person’s only such error.) But there are perhaps more difficult cases, in which it would be genuinely open to question whether adherence to some erroneous values constitued sufficient deviation from correctness to call the person’s sanity into question. These cases are quite likely to include the ones most pertinent – a person who makes no moral mistakes but places no value on their own continued existence, for instance, or who values being thin to the point of emaciation but is otherwise perfectly normal in his or her evaluative judgements. Here it seems that the decision on whether the mistaken belief in question is 24 ‘A person who, even on reflection, can’t see that having someone tortured because he failed to salute you is wrong plainly lacks the requisite ability.’ – Wolf, ‘Sanity and the Metaphysics of Responsibility’, op. cit., 145. 25 G.E. Anscombe, ‘Modern Moral Philosophy’, in R. Crisp and M. Slote, eds., Virtue Ethics (Oxford: Oxford University Press, 1997), 26–44, p. 42.
AUTONOMY, SANITY AND MORAL THEORY
53
evidence of insanity will itself involve a further value-judgement. Different people might take different views of the seriousness of the apparent impairment of the patient’s ability to form correct evaluative beliefs. This still leaves the most difficult point. In order to judge autonomy accurately, we shall need to know which moral theory is correct. Even if we accept that there is one correct moral theory, we may well not be wholly confident that we know what it is. In the example above, it is taken for granted that valuing being thin to a point danergous to life is a mistake, but may we not be in quite reasonable doubt about this? That would mean that, as well as there being room for doubt about the seriousness of the error, there is also room for doubt about whether it is an error at all. This will have the effect of producing considerable variation in the way in which autonomy is used as a means of deciding whether to respect patients’ wishes or not. Whether or not a person is judged to be autonomous will depend partly on whether they are thought to be able to want what they want to want, but also on whether the standards by which they judge their own wants are regarded as acceptable standards by whoever happens to be making the judgement as to whether or not they are autonomous. Thus a single person could be regarded as autonomous by one person or group of people and as non-autonomous by another, depending on whether or not they view her evaluative standards as reflecting truly the evaluative facts. The Reason View, when it comes to practical applications, is not a single view at all, but a multiplicity of different views varying as widely as do opinions on what is good and what is bad. If variation in moral opinion is less great than it is sometimes supposed to be, then perhaps the number of different ‘Applied Reason Views’ will be relatively small. Nonetheless, the central fact remains that if I am correct about the shortcomings of other accounts of autonomy, then the very notion of autonomy is one to which content can be given only from within a wider meta-ethical and moral position. Whether or not someone is autonomous will inevitably depend on the values that they have, and on who is judging the question. Any aspiration to make the issue of whether or not someone possesses autonomy one which could be settled from some sort of neutral perspective will have to be abandoned. With that in mind, it is time that we looked at the questions raised at the start of this paper. What prospect is there that this conception of autonomy could justify our thinking that the fact that a person is autonomous means that it would be impermissible, for example, to give her medical treatment contrary to her informed wishes? And is there any prospect of it being
54
IAIN LAW
possible to build a moral theory on the basis of such a conception of autonomy as this? The first question may be put the following way. Recall again the case in which I have to decide whether or not to treat against the patient’s wishes. Why should my belief that the person is autonomous give me a reason not to treat? Why should I not say, ‘So she’s autonomous. So what?’ I think that, equipped with the view of autonomy defended here, we can say this: if someone is demonstrated to you to be autonomous, that entails that you are satisfied that this individual is capable of discerning and responding to what you take to be the cognitive and normative truth. There is no possibility of someone whom you judge to be autonomous then turning out not to be amenable to reason, or acting on the basis of an inability to see what is right or wrong, or good or bad. A judgement that a person is autonomous means partly that you share to some extent an outlook on the world with that person. You will not necessarily share all the beliefs and values of the other, but you will judge that, however their beliefs and values may differ from your own, they do not stray into the realms of the unacceptable. At some level, you accept that they are not wrong (either incorrect or morally wrong) to want to act as they do and to want to live the sort of life that they do. Apprehending the existence of this sort of shared outlook on life might be thought to give us an excellent reason to think that we should not interfere with each other in the pursuit of our ends. Exactly how much of a good reason it gives us may depend on how closely you require another person’s basic outlook on life to match yours for you to consider them sane. If your point of view requires very close correspondence then your reasons for regarding those whom you regard as sane as inviolable will be stronger. If, on the other hand, your requirements for how much ability to grasp the true and the good another person must have in order to be sane are fairly relaxed, then the discovery that you judge someone to be sane may carry with it a less strong reason to accept that she is entitled not to be interfered with even for her own good. It seems that the fewer people one regards as autonomous, the more likely one is to see them as beings with whom it would be impermissible to interfere, and the more people one sees as autonomous, the less reason one will see so to refrain. So the fact that a person is autonomous gives good grounds for thinking it wrong to treat them against their will. What is now less clear, it seems to me, is whether finding that someone is not autonomous gives good grounds for thinking it permissible to treat them against their will. If I am aware that my judgement that the patient is not autonomous rests upon contest-
AUTONOMY, SANITY AND MORAL THEORY
55
able grounds, that must surely make me less comfortable in going against my patient’s wishes. While I may think that the patient exhibits beliefs so wrong that they must lack the capacity to form correct beliefs, my awareness of reasonable disagreement with my judgement is likely to make me feel uncomfortable. Few of us, I think, are so sure of our value-judgements that being aware of the value-laden nature of judgements about autonomy would leave us undisturbed. Believing that judgements about autonomy are value-neutral can make it easier to go against people’s wishes, at least when we are making these judgements on the basis of underlying value-judgements without realising it.26 Being aware that those judgements are value-laden will, quite properly I suggest, make it harder. Given the quite reasonable doubts most of us have about our moral views, the notion of autonomy I advocate here means that lack of autonomy should only very cautiously, if at all, be used to justify disregarding people’s wishes.27 Turning to the second question, it might be thought that any attempt to make use of the notion of autonomy to ground a moral theory will have to get over the objection that it is trying to get an ‘ought’ from an ‘is’. This is not an objection which applies to this case, however. In fact, it is very clear that far from being able to stand as the foundation of a moral theory, the conception of autonomy defended here is complete only if it is conjoined with some moral theory or other. It should be no surprise that it is possible to generate moral principles from various versions of this view of autonomy, since each of them necessarily presupposes certain moral truths. We have ceased to be trying to get an ‘ought’ from an ‘is’, and are now just getting one ‘ought’ from another. In other words, it is not possible to derive moral theories from the conception of autonomy I have defended, since that conception itself depends on the acceptance of moral truths. Any attempt to do so would always be vulnerable to accusations of vicious circularity. To conclude, then, I have tried to show some of the consequences of adopting what is, I would argue, the most defensible view of autonomy. 26 Possibly one example of this is the recently reported view of the Department of Health that no sane person would voluntarily and without pecuniary motive donate a kidney, while alive, to another. Those who do, like the group known as ‘Jesus Christians’, are regarded as having ‘psychiatric problems and so need[ing] to be protected from themselves’ – Jon Ronson, ‘Blood Sacrifice’, The Guardian Weekend, April 6th 2002, 19–23, p. 19. 27 This holds even more where the judgement about lack of autonomy is based on one about lack of sanity. It is still possible to make a value-neutral judgement that a person is not autonomous, if the ground of that judgement is, say, that they cannot control their first-order desires.
56
IAIN LAW
The most significant of these consequences is that the concept of autonomy cannot serve the function which many have hoped it could, namely to ground and justify our moral judgements and theories.28 Department of Philosophy University of Birmingham Edgbaston Birmingham, B15 2TT UK E-mail:
[email protected]
28 Thanks to Nick Dent, Brad Hooker, Mark Walker and to an anonymous referee for
Res Publica for their comments.