Relational Autonomy and Perfectionism

Joseph Raz’s The Morality of Freedom (1986) is well known for defending both a perfectionist form of liberalism and an ‘externalist’ conception of autonomy. Raz says that ‘the ideal of personal autonomy is the vision of people controlling, to some degree, their own destiny, fashioning it through successive decisions throughout their lives’ (Raz 1986, p. 369). On his account, autonomy is an important component of human flourishing – it is a moral ideal – and external social conditions, particularly the presence of adequate options, are necessary conditions of living an autonomous life. Since Raz also defends perfectionism, the position that the state should promote sound moral ideals or conceptions of justice, it is incumbent upon the state to identify and provide the minimally adequate options necessary for citizens to secure autonomous and flourishing lives.

Raz’s approach is congenial to feminist or ‘relational’ conceptions of autonomy. Feminists as early as Mary Wollstonecraft in her Vindication of the Rights of Woman pointed out that – because of extremely limited social opportunities – the ideal of autonomy was not typically accessible to women (Wollstonecraft 1792). This suggested to Wollstonecraft and others that an autonomous life requires significant social scaffolding such as (at least) economic independence, equal education for women, citizenship, civil rights, and so forth (Mackenzie 2017). ^[1] Contemporary feminist philosophers who study autonomy have also been preoccupied with the relationship between the social and personal autonomy (e. g. Mackenzie and Stoljar 2000). In particular, does the social scaffolding that seems necessary for autonomy play a causal or a constitutive role in securing an autonomous life? The former approach is exemplified in Wollstonecraft’s own view. She thought that women were just as capable of ‘independence of mind’ and virtue as were men, but that social conditions limited their ability to cultivate such virtue. As Catriona Mackenzie argues, Wollstonecraft was ‘an early relational autonomy theorist’ due to her emphasis on the social conditions for autonomy (Mackenzie 2017). But the notion of autonomy as independence of mind was the equivalent of an internalist condition of autonomy. On such internalist views, social conditions promote or hamper the psychological states that are necessary and sufficient for leading an autonomous life, but they are not in principle inconsistent with autonomy.

However, on externalist views of autonomy, such as that of Raz, the conditions required for autonomy play a constitutive as well as a causal role in securing an autonomous life. Raz offers three necessary conditions of an autonomous life: mental abilities, freedom from manipulation and coercion by others, and adequate external options. Through the examples of The Man in the Pit and The Hounded Woman, Raz argues that, to be adequate, options must be non-trivial and of sufficient variety (Raz 1986, pp. 373f.). The Man in the Pit can make choices such as when to move around in the pit, what time to wake up and go to sleep, and so forth. But these are all ‘short-term and negligible’ and are not adequate for an autonomous life (Raz 1986, p. 272). The Hounded Woman is on a desert island with a carnivorous wild animal that is hunting her down. She has some options, including non-trivial ones, but she is consumed by the need to protect herself from the beast. An inadequate range of options prevents her from living an autonomous life. These requirements imply that even if women (or others) have access to education that cultivates their reason and virtue, they will still lack autonomy when their options are trivial or lack variety. If women are limited to for instance sitting quietly doing needlework, or if they have only two feasible life plans open to them – nun or wife – their options will be inadequate and their autonomy will be compromised.

Raz’s notion of autonomy is externalist and constitutively relational. He is also a proponent of perfectionist liberalism. ^[2] But he does not (as far as I know) claim that his account of autonomy entails perfectionist liberalism. For Raz, perfectionism would be correct even if autonomy were not a moral ideal, and the fact that autonomy in particular is a moral ideal (and conceptually requires adequate external options) does not itself entail perfectionist liberalism. John Christman however proposes a logical connection between the two theses and claims that externalist accounts of autonomy should be rejected on the basis that they are perfectionist (e. g. Christman 2009, p. 173; Christman 2004). Christman advocates political liberalism and hence any doctrine that entails perfectionist liberalism is false. The structure of this argument – which I will call Christman’s perfectionism argument – is as follows:

1. Externalist (constitutively relational) theories of autonomy imply perfectionism.

2. Perfectionism is not defensible.

3. Therefore, externalist (constitutively relational) theories of autonomy are not defensible.

In the first section, I examine the meaning of ‘perfectionism.’ I argue that the argument equivocates over the meaning of ‘perfectionism.’ Once we disambiguate ‘perfectionism,’ premise (i) is either true but uncontroversial or false. That is, the implication from externalist theories of autonomy to perfectionism can be called into question. Moreover, premise (ii) presupposes a disambiguation of ‘perfectionism’ under which (i) is false. Hence the argument fails because premise (i) is true only if the meaning of ‘perfectionism’ in premise (i) is different from the meaning of ‘perfectionism’ in premise (ii). In section two, I turn to the debate between political liberalism and perfectionist liberalism. Is perfectionism flawed and indefensible? I limit myself to examining what is perhaps the most important current argument against perfectionism, a respect argument. ^[3] Like Christman, Martha Nussbaum defends political liberalism (Nussbaum 2011). Nussbaum claims that the costs of perfectionism are too great; if the state is permitted to prioritize sound conceptions of the good over unsound ones, perfectionism will inevitably disrespect the persons who hold unsound but reasonable conceptions. ^[4] I argue that critics of perfectionist liberalism misunderstand what is required to secure respect: respect is secured through certain processes and does not require that the outcomes of processes, namely the conceptions of the good that are endorsed by particular agents, are themselves respected. When respect is understood this way, perfectionist liberalism is not subject to the respect argument. In the concluding section, I briefly address a further argument against constitutively relational theories of autonomy that is advanced in Christman’s discussion (Christman 2009). He implicitly suggests that there is no theoretical need to posit externalist theories of autonomy. Rather, we should adopt a procedural theory of autonomy that is morally-neutral and useful to advance political liberalism. I claim however that the procedural conception of autonomy that Christman defends is not morally-neutral; it cannot do without the social scaffolding that Wollstonecraft and others have pointed out is necessary to the autonomy of women and other historically oppressed groups. In short: even Christman’s procedural account potentially suffers from the perfectionism he attributes to externalist theories. ^[5] It has no advantage over externalist theories on that score.

1 Christman’s perfectionism argument

John Christman defines externalist accounts of autonomy as theories in which features of the ‘interpersonal or social environment’ are included in the ‘defining conditions’ of autonomy (Christman 2009, p. 166). He has no objection to requiring options in one sense; he says that having a range of options is ‘generally required for all life pursuits’ (Christman 2009, p. 170). Rather, he objects to the way in which externalist theories build moral criteria into the conditions or ‘adequate options’ that are required for autonomy. The first premise of Christman’s perfectionism argument asserts that externalist conceptions of autonomy imply perfectionism. In this section, I evaluate this premise and argue that we first need to disentangle three possible perfectionist theses. Under one disambiguation of ‘perfectionism,’ premise (i) is true but uncontroversial. On another disambiguation – the one required for premise (ii) – premise (i) is false.

Let us first consider the ways in which externalist theories build in moral conditions. On Raz’s account, the notion of adequate options is delivered by his pluralism about the objective goods required for human flourishing. To lead autonomous lives, people have to have the ability to select among and pursue a variety of objective goods. Christman’s argument also explicitly targets another externalist account, that of Marina Oshana (Oshana 2006; see Christman 2009, pp. 167ff.). For Raz, morality is built into the notion of adequate options rather than independence; independence requires freedom from manipulation and coercion and can be understood as a non-moral criterion. However, for Oshana, independence is a moral notion. On her view, agents who are subject to ‘domination’ or the possibility of arbitrary interference in their daily lives do not have ‘substantive’ independence and hence cannot be autonomous. ^[6] For example, suppose young African-American men are routinely subjected to police profiling or arbitrary stop and search policies. Even when they are not stopped by the police, and therefore not subject to actual interference or coercion, they are not secure from possible interference. Thus, they lack the substantive independence or ‘practical control’ over their daily lives that is required for an autonomous life on Oshana’s account. For the sake of argument, we will assume that the possibility of this kind of arbitrary interference exists under nonegalitarian or hierarchical social structures such as those perpetuated by racism and sexism. ^[7] Oshana’s notion of substantive independence therefore seems to presuppose a commitment to a substantive egalitarian position.

Does Raz’s or Oshana’s conception of autonomy imply perfectionism? Christman comments that these theories imply that ‘having certain opportunities is more objectively valuable, as determined philosophically by theorists, independently of what the agent wants’ (Christman 2009, p. 170). Moreover, on externalist theories, ‘values and moral principles can be valid for a person independent of her judgment of those values and principles’ and ‘there are certain objectively determined intrinsic values that should guide individual and social action independently of the endorsement of those values by minimally rational, autonomous individuals’ (Christman 2009, p. 173). In these quotes, Christman is pointing to one possible disambiguation of ‘perfectionism’ – one that is presupposed by all perfectionist theories, namely the thesis there is an ‘objective account of the good’ (Wall 2012). For instance, ‘human nature perfectionism … give[s] an account of the properties or capacities that are central to human nature and the development of which have value’ (Wall 2012). These properties and capacities are objectively good, not good ‘in virtue of the fact that they are desired or enjoyed by human beings’ (Wall 2012).

However, this first disambiguation should be distinguished from two closely related theses, ethical perfectionism and political perfectionism. The former is ‘a moral theory [that] directs human beings to protect and promote objectively good human lives’ (Wall 2012); the latter is the position that ‘other things being equal, one should favor political institutions and state policies that do the best job of promoting the good in the context in which they apply’ (Wall 2012). The third thesis is the key to the dispute between political liberalism and perfectionist liberalism. As Raz puts is, political liberalism endorses whereas perfectionist liberalism denies a ‘doctrine of restraint [which] advocates neutrality between valid and invalid ideals of the good’ (Raz 1986, p. 110). On perfectionist liberalism, the state is entitled to ‘help acceptable ideals more than unacceptable ones’ and to hinder unacceptable ideals more than acceptable ones (Raz 1986, p. 111). It is the latter thesis that is employed in premise (ii) of the perfectionism argument.

Externalist conceptions of autonomy may seem imply the first perfectionist thesis, namely that there are objective goods. Raz’s theory includes a commitment to pluralism about objective goods; to be autonomous, agents must have access to adequate options in the form of a minimal number of these objective goods. However, it should be noted that both Raz and Oshana are also engaged in a conceptual project of identifying the necessary conditions of an autonomous life. This conceptual project does not itself imply the first perfectionist thesis. For instance, Christman thinks that Oshana’s notion of substantive independence is committed to egalitarianism; it is therefore not morally neutral. But even if this is true, it does not establish that Oshana’s conceptual analysis entails that egalitarianism is objectively correct. Her view could be articulated as a conditional that does not take a stand on whether either autonomy or the necessary conditions of autonomy are objective goods, such as: ‘if we are committed to autonomy, we must also be committed to some form of egalitarianism because only the latter can secure the substantive independence or practical control this is required for autonomy.’ Raz’s theory also provides a conceptual account of the ingredients required for an autonomous life – recall the examples of the Hounded Woman and the Man in the Pit. It may not be conceptually necessary that the external options required for an autonomous life correspond to objective goods, even on his account. Hence, the question of whether there are objective goods seems independent of the conceptual relationship between the notion of autonomy and external moral conditions.

I will accept for the sake of argument that externalist theories entail or include the first perfectionist thesis. If perfectionism is understood to mean a commitment to objective goods, then premise (i) of Christman’s argument is true. It is also (relatively) uncontroversial. A commitment to objective goods, sound moral ideals and so forth, is tantamount to a rejection of moral relativism in general and relativism about the value of autonomy in particular. All theories of autonomy, internalist and externalist alike, presumably are committed to the value of autonomy; the claim that autonomy is valuable is a theoretical posit that is not dependent on people desiring or enjoying autonomy. It is a separate question however whether externalist theories of autonomy and a concomitant commitment to objective goods imply either ethical or political versions of perfectionism. Christman’s discussion implies that a commitment to objective values entails the further claim that ‘state policy should be guided by those objective values’ (Christman 2009, p. 173). But, as Raz notes, although it may have a ‘paradoxical air’ to claim both that there are sound moral ideals or objective human goods, and that there is no moral or political obligation to promote these goods, it is not logically inconsistent (Raz 1986, p. 111). Indeed, the doctrine of political liberalism that Christman endorses adopts precisely this strategy. Political liberalism says that even if there are objective conceptions of justice or sound moral ideals (that is, even if the first perfectionist thesis is true), it does not follow that the state is permitted to promote such objective goods. One the contrary, the state should adopt a doctrine of restraint and remain neutral in the face of disagreement among reasonable conceptions of the good, some of which correspond to objective conceptions and some of which do not. Hence the claim of premise (i), once ‘perfectionism’ is disambiguated, is either uncontroversial, stating that externalist theories of autonomy deny moral relativism, or false because externalist theories of autonomy – despite being committed to objective goods – do not imply political (or ethical) perfectionism.

2 A respect argument against perfectionist liberalism

Let us now turn to a consideration of premise (ii), the assertion that perfectionism is not defensible. For the purposes of this premise, perfectionism corresponds to political perfectionism (which Nussbaum calls ‘perfectionist liberalism’). One of the most influential current arguments against perfectionist liberalism is that it is inconsistent with respect. It is alleged that perfectionist liberalism – the doctrine that it is permissible for the state to take sides on questions of justice and objective goods – is incompatible with appropriate respect of citizens who have reasonable disagreements with the state (e. g. Nussbaum 2011). I claim that this argument misconceives what is required for respect. It assumes that respect requires remaining neutral about these competing conceptions of the good; that is, that respect is secured if and only if the state remains neutral with respect to the outcomes of each citizen’s reasoning process. On the contrary, respect is inherent in the processes or procedures implemented between the state and citizens affected by the decision. Respect is not tied to a decision-maker’s or the state’s attitude to the outcome of these processes.

As explained above, perfectionist liberalism should not be confused with the claim that there are objective goods, sound moral ideals or objective answers to the question of what conditions best promote human flourishing. Indeed, without first rejecting moral relativism the dispute between political liberalism and perfectionist liberalism would not get off the ground. The dispute arises when it is assumed that there are sound moral ideals: the question is, given that there are sound moral ideals, is perfectionism defensible? As Steven Wall points out, assuming that there are objective goods and sound moral ideals, the doctrine of political liberalism comes with considerable costs (Wall 2014, pp. 472f.). Political liberalism distinguishes between reasonable or legitimate conceptions of the good and sound or just ones. It says that ‘any conception of justice that could be reasonably rejected by a member of [the group of reasonable citizens] must not be enforced’ even if it is sound (Wall 2014, p. 471). Wall further points out that political liberalism is potentially costly because ‘[i]f one thought, in addition, that the set of conceptions of justice that could be legitimately enforced in a political society did not include the correct or best conception of justice, then one would need to conclude that legitimacy obstructs the realization of justice’ (Wall 2014, p. 471). He also observes that the weaker one’s conception of reasonableness, the more likely that a sound moral ideal will be illegitimate, and the greater the risk that the doctrine of political liberalism will exclude sound conceptions of the good. For example, Nussbaum adopts a permissive conception of reasonableness under which a conception of the good can be reasonable even if it is epistemically flawed or inconsistent, such as astrology or certain religious doctrines (Nussbaum 2011, pp. 24–28). Thus it seems that the onus is on proponents of political liberalism to show why we should accept that cost.

Nussbaum’s respect argument is meant to tilt the scales in favor of political liberalism by showing that perfectionist liberalism would incur a greater cost of undermining respect (Wall 2014). The basic idea is that permitting the state to select certain reasonable conceptions of the good for favorable treatment would be disrespectful to the people who hold reasonable conceptions that are not favored. A corollary is that it is not disrespectful to people whose conceptions of the good are unreasonable to favour the reasonable conceptions over the unreasonable ones. Nussbaum states her argument as follows: ‘Respect is for persons, not directly for the doctrines they hold, and yet respect for persons leads to the conclusion that they ought to have liberty to pursue commitments that lie at the core of their identity, provided that they do not violate the rights of others and that no other compelling state interest intervenes’ (Nussbaum 2011, p. 17; cf.; Wall 2014, p. 477). In other words, Nussbaum acknowledges that respect for persons does not require ‘direct’ respect for the content of their conceptions of the good. It is because certain conceptions of the good are ‘identity-constituting’ for certain people, that respecting the person requires respecting the content of their conceptions of the good.

I propose that Nussbaum misconceives the way in which respect for persons by the state is and should be secured. Respect does not require that the state remain neutral or refrain from overriding the contents of people’s identity-constituting conceptions of the good. Rather, respect is secured when certain processes are in place. To clarify the argument, I briefly sketch two examples: one of collective decision-making and a second of authoritarian decision-making.

Consider first collective decision-making, such as decision-making in any ordinary workplace committee. In such committees, there is a requirement that the decision-maker (the collective or committee itself) respect all members including those who disagree with the committee’s final decision. But respect of the members making up the committee does not imply that the committee is required to maintain neutrality among all the reasonable points of view expressed on the committee. Respect of persons is secured, not by respecting the contents of their points of view, but rather respecting that they are rational agents with a point of view. As Jeremy Waldron puts it, treating with respect ‘involves paying attention to a point of view and respecting the personality of the entity one is dealing with’ (Waldron 2011, p. 16). Respect is a matter of how the participants in the process of collective decision-making are treated: they must be ‘treated as beings capable of explaining themselves,’ given the opportunity to express the reasons for their point of view, and given the opportunity to respond to counterarguments disagreeing with their positions (Waldron 2011, p. 16). If democratic decision-making is collective decision-making on a larger scale, and the state’s decision is analogous to that of a committee, it would be absurd to think that a state is required to remain neutral with respect to a citizen’s substantive views.

A second example reinforces the point. Consider decision-making of a judge in a court. This is not the product of collective decision-making, but rather an act of adjudication among different ‘points of view.’ A court judgment cannot remain neutral among these points of view. But this does not mean that the court decision disrespects the parties with different points of view from that of the court decision. Waldron describes how to ensure that the different parties to a dispute are respected:

… the evidence is made available to be examined and confronted by the other party in open court. And each party has the opportunity to present arguments and submissions at the end of this process and to answer those of the other party … they are listened to by a tribunal that … is bound in some manner to attend to the evidence presented and to respond to the submissions that are made in the reasons it eventually gives for its decision.. [These characteristics] capture … that law is a mode of governing people that treats them with respect. (Waldron 2011, pp. 16f.)

Most importantly, parties must be treated as rational agents who are answerable and ‘capable of explaining themselves’ (Waldron 2011, p. 16).

Denise Meyerson has recently pointed out that social psychological evidence supports a process-related conception of respect. People do not value processes only to the extent that the processes achieve the desired outcomes or feel respected only when a decision implements their own point of view (Meyerson 2015, 254 ff.). ^[8] People care about interpersonal aspects of processes, such as whether decision-makers are polite and trustworthy and whether processes are fair (Meyerson 2015, p. 256). A crucial condition for promoting respect is providing the opportunity to be heard:

The opportunity to present one’s own perspective is perceived as a symbolic marker of inclusion within the larger community, conveying the information that one is a valued member of the group, whose views are worthy of being heard. Conversely, not to be allowed to present one’s own side would be experienced as an expression of contempt, which in turn would detrimentally affect one’s sense of self-worth or respect. (Meyerson 2015, p. 259)

Broadly, the decisions of states can be understood as corresponding to the model of decision-making in either of the two examples above. Either the state represents the collective and its decisions are a product of collective decision-making; or, the state is akin to a quasi-authoritarian decision-maker who, especially in formulating policy, has to adjudicate among competing interests. In neither case however does respect require that the decision-maker or the state remain neutral with respect to the reasonable disagreements of its citizens.

Nussbaum might respond that treating someone as a rational agent with a point of view itself requires respect for their conceptions of the good or comprehensive doctrines. She writes that we ‘respect persons [when] we think that their comprehensive doctrines deserve space to unfold themselves, and deserve respectful, nonderogatory treatment from government’ (Nussbaum 2011, p. 33; cf. Wall 2014, p. 477). The process-related respect I have described precisely captures the first requirement: in giving people the opportunity to explain themselves, or the opportunity to be heard, we allow their conceptions of the good ‘to unfold.’ But this does not entail the further step that it is unjustified to impede the exercise of such conceptions. Moreover, Nussbaum’s argument linking respect of a person’s conception of the good to respect for her identity has an overlooked, unpalatable consequence. We can accept that certain conceptions of the good are identity-constituting. Further, on Nussbaum’s account, because political liberalism adopts a distinction between reasonable and unreasonable conceptions of the good, there is no requirement that the state not hinder the pursuit of unreasonable conceptions. Presumably the state can prohibit white supremacist anti-miscegenation laws. There is no reason to suppose however that such unreasonable conceptions of the good could not themselves be identity-constituting. It is plausible that white supremacists hold an identity-constituting, yet racist and unreasonable, conception of the good. But if the very act of hindering such views is disrespectful of the person’s identity or rational agency, it follows that the state is permitted to disrespect persons who hold unreasonable conceptions of the good. The unpalatable consequence for political liberalism is that persons holding unreasonable conceptions of the good are not entitled to respect. Yet, despite the unreasonable, racist views of a white supremacist, it does not follow that the person, qua human being and rational agent, should not be respected. The process-related conception of respect explains how respect is possible in this situation and avoids the unpalatable consequence of Nussbaum’s account.

3 Internalist autonomy?

I have established that Christman’s argument against externalist conceptions of autonomy is unsuccessful. Even if it is conceded that Raz’s and Oshana’s externalism about autonomy implies perfectionism, or includes the position that autonomy is an objective good, it implies only the first perfectionist thesis, namely the position that there are objective goods. It does not imply the political perfectionism that Christman and other proponents of political liberalism want to reject. Moreover, even if we assume for the sake of argument that premise (i) of the perfectionism argument is true under all disambiguations, and hence that externalist conceptions of autonomy do imply political perfectionism, premise (ii) cannot be maintained. Perfectionist liberalism is a defensible thesis that is not subject to Nussbaum’s respect argument. ^[9]

In this section, I briefly address a further point against externalist theories of autonomy that is implicit in Christman’s discussion. His project proposes in effect that we do not need to posit an externalist (constitutively relational) conception of autonomy because there is a better, morally-neutral, alternative. He argues for a morally-neutral, internalist and procedural conception of autonomy precisely because he thinks it compatible with political liberalism. I claim that Christman’s internalist view cannot avoid moral commitments and hence may itself imply the kind of perfectionism he wants to reject. ^[10]

First, let us return to one of the motivations of feminist, relational theories of autonomy, namely to explicate the multiple ways in which autonomy is impeded by the social conditions of oppression. Christman agrees with relational autonomy theorists that:

… in most cases of oppressive social relations, autonomy is squelched by those relations. But the particular reason why autonomy is lost here is that the victims in such relations are unable to see themselves (without alienation) as even part author of the social narrative of which they are a part. It is not social connectedness as such, or even the hierarchical or unequal social relations that disturb autonomy … (Christman 2009, p. 172; my emphasis).

On Christman’s procedural view, authenticity or nonalienation (conceived as an internal, psychological state) is a necessary condition of autonomy. If the practical identities of oppressed agents are alienated and inauthentic, they will lack autonomy. At first sight, this focus on internal nonalienation flies in the face of the feminist claim that external, social conditions are crucial for autonomy. There is a way to reconcile the two views however. Christman can acknowledge that social conditions are causally important in producing practical identities. His approach can accept that social scaffolding is required to bring about the critical reflection and nonalienation needed for autonomy. Thus, although he does not adopt an externalist and constitutively relational account, his own position can acknowledge the causal effect of oppressive external conditions on autonomy.

On the other hand, an acknowledgement that social conditions influence both the social narrative and the ways in which practical identities or self-conceptions are formed may be problematic for Christman’s account. First, the emphasis on nonalienation has a counterintuitive consequence. Christman says that it is not the oppressive script ‘as such’ that is incompatible with autonomy; rather, oppressed agents will fail to be autonomous because of the alienation they experience in attempting to reconcile the social narrative of oppression with their practical identities. Yet, practical identities often correspond to the social narrative; people’s self-conceptions often endorse the norms of dominant social scripts. For instance, women apply gender norms to themselves and evaluate their practical identities as women, mothers or wives (for instance) in the light of these norms. Rational agents living under oppression often are not alienated from oppressive social narratives precisely because it is the social narrative that forms their self-conception. Even if they do not explicitly endorse oppressive scripts, they adjust their behavior to accommodate them, and in so doing themselves reinforce the social narrative. In other words, rational agents take on oppressive scripts as part of their self-conceptions by either explicitly endorsing them or accommodating their behavior to them (cf. Stoljar 2015). If we assume that nonalienation is sufficient for autonomy, Christman’s position has a counterintuitive consequence: the many people whose practical identities correspond to the oppressive social narrative are authentic and autonomous.

Christman suggests that, although nonalienation is necessary for autonomy, it may not be sufficient. In addition, agents’ critical reflection must not be constrained by ‘reflection-distorting’ factors such as ‘being denied minimal education and exposure to alternatives’ (Christman 2009, pp. 147, 155). It is unclear precisely what these further conditions entail, but it is implausible that they could be morally-neutral. For most people, if critical reflection is to have a chance of dislodging the oppressive patriarchal narrative, institutions that provide equal education to girls must be in place. There must be a prior commitment to a conception of the good, namely that gender-neutral educational opportunities are morally required. Thus, Christman’s internalist conception will have to presuppose that ‘having certain opportunities is more objectively valuable, as determined philosophically by theorists, independently of what the agent wants’ (Christman 2009, p. 170). If so, the internalist account of autonomy is parallel to externalist accounts in the respect that certain moral conditions are required for autonomy to achieved. ^[11] Christman’s theory is therefore itself implicitly committed to at least the first form of perfectionism, namely that the social conditions required to bring about nonalienated (and nonoppressive) practical identities are objective goods; they are valuable independently of whether people desire these social conditions.

As we saw, perfectionism in the first sense does not imply political perfectionism, so it is a further question whether the perfectionism that is implicit in Christman’s position also undermines political liberalism. Notice however that if there is reasonable disagreement over whether the conditions that are conducive to the development of (internalist) autonomy are morally appropriate – for instance, disagreement over whether educational institutions must respect gender equality – then political liberalism would advocate restraint on the part of the state in implementing such conditions. (Political liberalism would have the unpalatable consequence that egalitarian public education is not permitted.) However, if political liberalism recommends restraint vis-à-vis the implementation of moral conditions required for autonomy, it may be self-defeating. One of the core values of liberalism, including political liberalism, is that of individual autonomy. It is difficult to see how the central tenet of political liberalism, namely that it would be impermissible for the state to promote the value of autonomy if doing so presupposes a contested conception of the good, could be reconciled with a central tenet of liberalism, namely the requirement that the state promote autonomy as a core value.

Thus, the problem for Christman’s internalist theory is twofold. First, if externalist conceptions of autonomy should be rejected because they are perfectionist (in the first sense), the internalist conception should be rejected on the same basis. Second, if Christman’s conception of autonomy is not morally neutral, it is unclear whether it is compatible with political liberalism. In principle, a political liberal like Christman could advocate a doctrine of restraint with respect to autonomy itself or to the conceptions of the good that must be implemented to promote autonomy. But restraint with respect to the value of autonomy seems incompatible with conceiving of autonomy as the central value of liberalism overall.