For all that it is a concept in common use in political philosophy, there is little consensus about what autonomy actually is. Clearly, this is a dangerous state of affairs if one is trying to justify a theory of liberalism based on a commitment to the value of autonomy. Justification is apt to be misunderstood, with different readers taking âautonomyâ to mean different (sometimes wildly different) things. Opponents of such a theory run the risk of attacking it on an understanding of autonomy different from that on which it is actually based; a parallel danger for me will be unwittingly playing upon equivocations to make justifications seem more plausible than they in fact are. These dangers cannot entirely be avoided. The best that can be done is briefly to survey the territory, then clear the undergrowth a little by clarifying what I shall mean by âautonomyâ in what follows, and why. This chapter deals with that task.
Before I start, it will be useful to make clear two distinctions between ways of understanding autonomy. First, we might distinguish between autonomy conceived of as a local propertyâthat is, as a property of a person at a particular time and perhaps in respect of particular decisions or actionsâand as a global property, meaning a property of a personâs life as a whole. Secondly, we can distinguish between different types of properties that one might take autonomy to be. Some conceptions take autonomy to be a condition that someone can be in, and others take it to be a capacity that one might possess, the possession of which is itself valuable (independently of its exercise, for example). These distinctions will prove significant later on, especially when I show how my conception of autonomy differs from the others I examine.
Despite the lack of consensus surrounding contemporary conceptions of autonomy, I suggest that we can identify a common thread running through them all. Following John Christman, I suggest that the core concept of autonomy is something like self-governance, or control of oneâs commitments.1 Each of the conceptions autonomy that I examine below is a different understanding of a concept of self-governance.
There are three dominant conceptions of autonomy currently in circulation. First, there are conceptions of autonomy as rational self-legislation, most famously (though not exclusively) espoused by Kant and his intellectual successors. Secondly, there are conceptions that rely upon the notion of a hierarchy of motivations advanced by Harry Frankfurt and by Gerald Dworkin. Finally, there are conceptions built on a notion of self-authorship or individuality, such as that employed by Joseph Raz. Considering these three conceptions separately should not imply that they are necessarily in direct conflict with each other. As we shall see, I adopt my conception of autonomy partially in response to the flaws of the other positions, but I do not assume from the outset that it is impossible that the three families of conceptions examined here might coexist. There may just be three different and compatible ideals at work, each indicated by a different conception of autonomy.
1.1 AUTONOMY AND REASON
One family of substantive conceptions of autonomy understands self-governance to consist in rational self-legislation or determination. So, on these views, the concept of autonomy is to be fleshed out as an ideal of a condition possessed by agents at particular times: for an agent to be autonomous is for them to act with self-control in accordance with reason. Such conceptions fall into two main categories, which I shall call Kantian and non-Kantian rationalism respectively.
According to Kant, a person is autonomous if his will is self-ruling.2 To be self-ruling, the will must be âefficient independently of alien causes determining it,â where alien causes include not only the agency of others (as, for example, if one were being controlled by hypnosis) but also the laws of nature. The latter includes things like desires and impulses: such mental phenomena are part of the causal chain of nature, and therefore constitute alien causes. Kant also wanted to exclude lawless action, for that too is a sort of slavery (though to chaos, rather than to alien causes). So, the autonomous will must be in accordance with laws of a special kind, namely those that have as their source nothing other than the will itself. That, Kant said, meant behaving in accordance with the Categorical Imperative, for that is the only law that can be derived just from pure reason. To be autonomous, the will must not act on anything that might be an alien cause. Alien causes derive from the particular, contingent circumstances of individuals: desires, impulses, adverse conditions and pressures, and so on. To avoid these, the will must act in a way completely undetermined by the particular circumstances under which it is acting. That means, though, acting on motives that one might will that anyoneâincluding those who donât share oneâs particular circumstancesâ could also will. So, in Kantâs theory, the autonomous will is just one that acts in accordance with the Categorical Imperative.
Kantâs theory, as sketched here, has two components. First, there is the concept of autonomy for which he argued: autonomy is a property of a personâs will which consists in that will being determined only by laws given by the will itself. Secondly, there is the particular conception of autonomy that comes from his claim that the only law so given is the Categorical Imperative. It would be possible for someone to endorse the first component, but not the second. That is, we could agree that autonomy consists in acting only on laws given by the will itself, while thinking that the Categorical Imperative is not the only law that can be so given. We might, for example, disagree with Kant that all contingent factors count as alien causes in the relevant sense, and discriminate between different types of desire. Perhaps certain types of desire might not be alien to us, and that acting on those desires would not vitiate autonomy. Or, we might agree that all desires are alien causes, being part of the natural chain of causation, but think that there is nothing to stop the will from choosing autonomously to take them into consideration in its deliberation.
This gives us various possible Kantian conceptions of autonomy. The strictest position is that held by Kant himself: autonomy is equivalent to the behaviour demanded by morality, which is to say acting in accordance with the Categorical Imperative. Conceptions can vary in strictness by allowing that there might be other things than the Categorical Imperative that are non-alien in the relevant sense.
Contemporary philosophers who have used a Kantian conception of autonomy vary in their strictness. Jonathan Jacobs, for example, takes a position at the stricter end of the spectrum: he argues that someone who acts freely and voluntarily might nevertheless fall short of autonomy because they might fail to follow the moral law.3 Others defend conceptions that are less strict, sometimes considerably so. David Richards, for example, takes himself still to be a Kantian, but says that the requirement that the will is not determined by alien causes is satisfied by a weak endorsement condition: it is sufficient for autonomy to have exercised âthe higher-order capacity of the agent to choose her or his ends, whatever they areâ for those ends to count as non-alien.4 Richardsâ variant of the Kantian conception brings it close to collapsing into the conceptions I examine in Section 1.2, which emphasise higher-order endorsement as the key to autonomy. However, Richardsâ view remains distinct insofar asâlike Kantâhe takes the non-alienness of causes to be centrally important, and endorsement important as the means to guaranteeing non-alienness.
At this point, we might worry whether Richards, Jacobs and Kant are really all using the same conception of autonomy. Indeed, we might even worry whether a Kantian conception of autonomy can properly be applied to political philosophy at all, since in Kantâs own theory its role is just to identify the quality that the will must have if it is to be able to act in accordance with morality. Kantâs conception of autonomy is sometimes called moral autonomy, and is contrasted with personal autonomy (which is normally equated with one of the other conceptions of autonomy I examine in this chapter). Drawing this distinction is often intended to motivate our avoiding moral autonomy as a potential subject for political philosophers to consider. This sometimes seems to be the position of John Rawls, for example.5 Even if we grant that a Kantian conception of autonomy can play a role in political philosophy, there is dispute about whether the people who take themselves to be Kantians are working within such a framework. Onora OâNeill argues, of positions such as Richardsâ, that they badly misrepresent their own foundations if they claim Kantian support.6
These are interesting problems, and they have received some attention in recent literature.7 However, they are somewhat tangential to our present business. As a matter of fact, there are political philosophers who adopt a Kantian conception of autonomy and apply it to political philosophy. So, they are worth examining here, even if we worry about their Kantian pedigree. To put it another way, regardless of whether we think them genuine interpretations of Kant, Kant-inspired, or only pseudo-Kantian, they all share the core thought that autonomy consists in acting only from non-alien causes, which they all identify roughly as acting only in accordance with reason. Another group of conceptions of autonomy exist, which while they do not trace a lineage to Kant, still place emphasis upon the notion of rational self-legislation or direction. So, for example, Robert Ladenson sketches a conceptionâwhich he attributes to Deweyâon which autonomy is identified with a life lived in accordance with reason, understood as
These âmore developedâ capacities are contrasted with lower ones such as emotions, instincts and habits. Hence, we suppose, for Ladenson being autonomous consists in being guided by reason and eschewing those lower capacities. Ladensonâs ideal has a teleological airâthat is, it identifies an end-state condition which is valuable, and identifies autonomy with what is needed to get to it. Other proponents of rationalist autonomy avoid this teleological character and instead take the value of autonomy to be the exercise of a capacity. Some writers on Stoicism, for example, take autonomy to be a condition consisting of control over oneâs urges and impulses, either through the exercise of willpower,9 or by the use of the capacity for reason.10 Others use a similar conception of autonomy, though they donât identify it as especially Stoic in nature. So, George Sher says that an autonomous agent is self-governing in the sense of âexercising their will on the basis of good reasonsâ;11 Keith Lehrer suggests that autonomy consists just in being governed by reason;12 John Benson says that to be autonomous one must âput oneself in the best position to answer for the reliability of oneâs beliefsâ;13 and Lawrence Haworth says that to be autonomous an agent must display âfull rationalityâ, which consists in both means-end reasoning and exercise of critical appraisal of the ends they adopt.14
In all these conceptions of autonomy there is a common emphasis upon rational behaviour. The concept of autonomy as self-governance is interpreted as an ideal of the exercise of reason. It is in respect of this emphasis on reason that the conceptions in this family are distinct from those I consider later. And because of this, these conceptions are all in danger of unjustifiably lionising reason over other possible sources of action and motivation. Why, we might ask, should we think that acting on reason is valuable, or at least the same sort of ideal as autonomy? Why should we think that only acting on reason can count as self-governance?
One consistent answerâbut one which we might be reluctant to acceptâ was given by Kant. Kant gave an objective criterion by which he judged that pure reason was the only non-alien cause. According to Kant we should distinguish between the phenomenal world, in which our desires are bound within a causal chain that destroys freedom, and the noumenal world, in which we are free from such causal determination. It is only insofar as we act without being determined by the tyranny of causation that we govern ourselves. So, Kantâs motivation for identifying autonomy with acting on pure reason flowed from his broader metaphysical theory, and it serves as a motivation only if we accept that metaphysics.
Suppose that we reject Kantian metaphysics, or at any rate want to avoid basing our political philosophy on such controversial territory. What then? I suggest that, with Kant out of the picture, there is no justification for placing such weight on the exercise of reason in our account of autonomy. After all, introspection suggests that there are other motivations which seem, as much as acting on reason does, to derive from the âselfâ. So, if our core concept of autonomy is that of self-governance, it is unclear why those other motivations mightnât also count. In the absence of a positive argument for abandoning common sense here (and assuming that we do not want to adopt the whole of Kantâs metaphysical position), this should make us wary of understanding autonomy in this way.
Admittedly, pointing out the absence of an argument for the emphasis on reason hardly refutes the rationalist conceptions of autonomy. However, there are other reasons to think that there is something unattractive about autonomy understood in this way. All the conceptions considered in this section take self-governance to mean self-governance according to what is demanded by reason. They differ in quite how stringent and specific they take those demands to be, but they are all committed to taking those demands to be binding. Hence, they recognise no authority for individuals over questions of what will make their lives go well. As we shall see in Section 1.3 I take this to a core component of the i...