Autonomy and cultural practices: The risk of double standards

The value of autonomy represents a barrier against unchecked paternalism, ¹ and this is a major reason to make autonomy a prominent liberal value. By implication, though, it would seem that non-autonomous conduct would symmetrically license paternalistic intervention. ² Even if such implication is rarely drawn directly, it is true that among the various arguments advanced for questioning toleration of cultural practices, one distinctive set focuses on the alleged lack of autonomy displayed by certain practices or cultural behavior. The claim is that people observing those practices or engaging in such behavior exhibit some defect in their personal autonomy regarding their values, attitudes, and choices. Examples vary from the much-contested practice of female (but also male) circumcision to the less dramatic, but no less controversial wearing of the Islamic headscarf, burqa or niqab. Non-autonomous conduct—so the argument goes—represents a form of harm to the self and to the whole group involved, reinforcing practices that are adaptive and self-abrogative, contributing to keep the agent, and the related group, in a family- or culture-dependent position that blocks her future opportunity for change and improving her condition. Hence non-autonomy and self-harm lead some liberals and feminists to invoke restrictions of the above cultural practices. ³ More generally, in many arguments doubting the legitimacy of toleration of certain conducts, the lack of autonomy judgments figure as important components, if not as the sole reason for limits of toleration.

In this paper I intend to dispute the lack of autonomy judgments on two different grounds: first on epistemological grounds, for I shall argue that no matter how autonomy is conceived, lack-of-autonomy judgments can rarely be drawn with confidence. Second, on grounds of political morality, for I shall argue that from a political point of view, drawing judgments on people's autonomy would threat the fundamental moral outlook of political liberalism. To this end, I shall first examine the assumptions and basis of the lack-of-autonomy approach; this analysis perforce leads me to unravel the notion of autonomy and its rival conceptions; eventually I shall be able to argue the epistemological difficulty to draw lack-of-autonomy judgments. I shall then proceed to examine autonomy from a political point of view. In this respect I shall take the suggestion by John Christman of autonomy as status marker of citizenship and expand it further. The view I shall defend is that liberal and democratic politics, though indebted to the value of personal autonomy in a fashion, requires only a purely political principle of autonomy which would precisely dispense with drawing moral distinctions among citizens on autonomy-based judgment. Finally, I shall draw the conclusions of my critique of lack-of autonomy judgments concerning the toleration of different cultural practices, and I will illustrate my point with some examples drawn from multicultural politics.

Lack-of-autonomy judgments

Autonomy conveys the idea of persons being able to regulate themselves according to ideals, values, and interests that can be said to be properly theirs. The concept of autonomy admits several interpretations, yet, no matter which conception of autonomy is preferred, it is far from clear that judgments on the autonomy of a certain conduct can be made from the perspective of an external observer. Judgments of this kind can be made with a certain ease when confronting cases of self-deception and akrasia, that is when the agent displays a failure of rationality concerning her beliefs, desires, and action. Akrasia and self-deception are typical cases in which the agent is not sovereign over her beliefs, desires, and actions, but is subject to some rebellious wish that drives her actions and beliefs beyond her control.

But leaving aside this type of lack of autonomy, and excluding coercion or direct manipulation, how can one judge someone to be acting non-autonomously if the action follows from an apparently free choice, and if the person appears to be in control and to know what she is doing? Which of the different conceptions of autonomy can support such a judgment and how?

Proceduralist approaches to autonomy

Inquiry into such questions has been deeply influenced by the seminal work of Harry Frankfurt ⁴ and Gerald Dworkin, ⁵ each of whom developed a procedural conception of autonomy (sometimes also called “hierarchical”;). ⁶ On this approach, autonomous action stems from autonomous volitions, which in turn are the outcome of a critical reflection on, and selection of, the first-order preferences that one happens to have. After critical reflection, a person's actions are directed by her own will, shaped after preferences that can be said to be authentically hers. The two features of the core idea of autonomy—self-determination and authenticity—are thus captured in this model. ⁷ The procedural model is both internalist and content free. By “internalist” I do not mean that “procedural independence” grants autonomy despite any external conditions such as manipulation, brainwashing, and coercion, but rather that from external conditions alone autonomy or non-autonomy cannot be inferred, while the analysis of second order reflection is crucial.

On this model it would seem prima facie impossible to make an external judgment of lack of autonomy, at least where the conduct under scrutiny is free of coercion and compulsion, and has passed a test of minimal rational competence, and so is neither akratic nor self-deceptive. Determining autonomous agency, on the proceduralist model, thus seems to be more a matter for critical self-knowledge and self-appraisal than for interpersonal relations. ⁸ In responding to the many objections it has elicited, however, the model has been refined and made more complicated, and, as a result, has ended up less purely procedural and internalist than in its initial specification.

The exclusive focus on internal conditions for critical reflection has seemingly made the model incapable of providing satisfactory accounts of both self-government and authenticity. Concerning self-government, a crucial objection has been raised regarding actions under duress or coercion. ⁹ Someone subject to these conditions—for instance with a robber's gun pointed at them—may want to behave reasonably in the circumstances, and his want maybe the result of a critical reflection that submission and handing the money over to the robber is the best course. Yet even though his action has been determined by his second-order volition, produced in a critical way, and thus seems to satisfy the conditions for autonomy on the hierarchical conception, his action can hardly be described as autonomous. The suggested solution by J.S. Taylor tries to account for the two contrasting moral intuitions concerning autonomy under duress: (a) that the agent did what he was asked following his second-order desire to be reasonable; and (b) that the agent's autonomy was nevertheless impaired because he had to cede a portion of his self-government. ¹⁰ The agent is autonomous with reference to his first-order desire, but if the satisfaction of his critical desire involves giving up his control of the situation, then it is a case of impaired autonomy with respect to the compliant action. Such a revision picks up the tension between the authenticity of the second-order desire and self-government and suggests that critical reflection in the Stoics' fashion cannot trump self-government, which conditions of duress impair beyond the critical capacities of agents.

Taylor's proposal seems fair enough, but imagine an analogous, less dramatic example of duress. Imagine someone raised in a very oppressive and authoritarian environment that has not impaired her critical capacities, but which makes it exceedingly costly for her to rebel against her family's rules. Like the man threatened at gunpoint, she may reflect on her first-order desire to escape and come to hold a second-order desire to be sensible and make the most of her lot, hence acts according to this critical desire. Is she acting autonomously? Here our intuitions are not as clear as in the gun-at-the-head case. If we accept Taylor's suggestion regarding the gun-at-the-head case, then we seem committed to accept it also in this more mundane case of duress and conclude that her autonomy is impaired. Thus the revised procedural model seems to admit of external as well as internal conditions, hence to allow for lack-of-autonomy judgment.

Yet, think of a variant of the previous example: a girl is raised in an oppressive and authoritarian environment, but she does not have the first-order desire to escape, and has contentedly shaped her first-order desire in keeping with her context, wholeheartedly endorsing that desire in her second-order reflection. In this case, we cannot say that she has relinquished her self-government, as she does not apparently experience any duress and is perfectly and happily integrated in her environment. Think now of a third possibility: her upbringing has not only affected her first-order desires but her capacity for critical reflection as well; it has been so thoroughly despotic that she has never learned to think for herself, to step back from her situation. In a way, it is a successful case of brainwashing. In this last case, she apparently lacks the competence condition for autonomy. What the three examples shows is the following: on the basis of the same view of autonomy, the very same outward conduct, in the same context can lead to different judgments as to the agent's autonomy depending on a variance in the agent's responsiveness to her environment.

In sum, the duress objection means that the procedural view faces a dilemma. Either we discount the issue of autonomy under duress and keep the model purely internal, or we account for duress in line with our moral intuitions, but thereby introduce external conditions, and so invite external judgments of lack of autonomy. While such external judgments can apparently be drawn, it remains that from the vantage point of an external observer, it is de facto impossible to differentiate the very different cases earlier, given that the context and the outward conduct are precisely the same and access to the agent's internal states is not granted.

Coming now to authenticity, here the pertinent objection to the hierarchical model questions whether critical reflection on one's first-order preferences is capable of granting the authenticity of second-order volitions without further conditions. The hierarchical approach actually is prone to an infinite regress and to incompleteness of authentication. ¹¹ If desires have an exogenous origin—notably, from socialization, culture, and genes—how can they ever be made authentic by hierarchical consideration? ¹² A first response invites the consideration of the history and etiology of desires as a key factor for authenticity; ¹³ alternatively, notions of endorsement, wholehearted acceptance, ¹⁴ or, at least, nonalienation from the desire ¹⁵ is said to authenticate one's desires. Yet none of these suggestions can properly solve the ab initio problem, i.e. the exogenous origin of preferences, which, after all, is not amenable to conclusive solution. Yet, historical analysis may actually reveal distorted factors in a person's desire formation such as manipulation, coercion, and conditioning that differ from the usual external influences and, at this juncture, again, the possibility of a lack-of-autonomy judgment remains available, though only in principle. The endorsement condition, in turn, captures the intuition that the autonomous person be endowed with some level of integration; however, it exposes the procedural model to another difficulty keenly detected by Marina Oshana. ¹⁶ If authenticity is the attitude of reflexive acceptance a person adopts toward her beliefs, values, and commitments, and, in general, her personal identity traits, it follows that if a person feels alienated from some of her traits, her autonomy is impaired according to the procedural model. But what if those traits are not a matter of choice but are “inescapable” such as gender, ethnicity, and sexual orientation? As Oshana notes, those traits cannot be denied without self-betrayal and self-deception by a reflective agent, but, nevertheless, the implications of those traits on the agent's choices and plans are often undesirable for the agent who wants to be self-ruled rather than role-ruled. Hence a certain alienation from some aspects of one's inescapable traits is needed in order to plan an autonomous life; and, in this respect, the reflective endorsement of inescapable aspects of one's identity cannot be a requirement for autonomy. Oshana actually stretches Frankfurt's notion of endorsement too far, given that he was thinking only of volitional states, and not of identity traits, as objects for authentication. Still, she has identified a weak spot here, because the reflective endorsement condition would seem to make autonomy more easily reachable by people situated in certain favorable contexts, where they experience no profiling as to certain collective characters, hence can endorse them without problem. Under this light, autonomy may look like a luxury for people with the right kind of background. ¹⁷ So once again we see that external elements influence a model which was meant to be purely internalist and procedural.

In sum, the procedural model scarcely manages to be self-contained. Whenever it intersects with external conditions, the possibility of an external judgment regarding the autonomous agency behind a certain conduct, choice, or lifestyle is in principle open, though de facto almost impossible to make, given the variances in how individuals internally respond to the same circumstances.

Substantive approaches to autonomy

The procedural view is, however, only one approach to autonomy; substantive accounts of autonomy also have been proposed. ¹⁸ The substantive idea of autonomy holds that it is insufficient that choices and actions are produced in accordance with a critical examination and endorsement of first-order preferences, independently of any normative constraints on the agent's self-regard and of the content of her choices. For one thing, some scholars claim that agents must meet certain substantive normative conditions in order to be autonomous, such as the capacity to speak for themselves, in their own voice, and to take ownership of their actions. According to this view, autonomy cannot simply consist of internal conditions, such as the identification with one's values, motivations, and actions, but must also include “agential authority,” which is implicit in the self-regard of the agent as a worthy member of the moral community, endowed with a voice and capable of answering for her actions. ¹⁹ This position amounts to a weak form of substantive autonomy. In contrast, fully substantive views require, on the one hand, normative restrictions on the kinds of preferences and values that confer autonomous agency to actions, ²⁰ and, on the other, that an adequate set of options be effectively present for autonomy to be practiced. ²¹ In other words, according to substantive views, autonomy depends on not just the structure, the history, and the coherence of values, beliefs, and volitions internal to the agent, but also on the normative adequacy of the internal set and, in some cases, the factual adequacy of the available options.

Leaving aside “factual adequacy” which deals with the circumstances for autonomy more than with autonomy itself, let me consider the criterion of “normative adequacy.” The latter can be specified as an adequacy of the agent as moral authority (weak-substantive-autonomy) or as an adequacy of the content of values and preferences that enable autonomous choices and forms of life (full substantive autonomy). The intuition here is that certain choices (a) produced by oppressive socialization, and (b) running against the agent's independence cannot count as autonomous. For example, choices in favor of self-enslavement and submission, regardless of whether they result from procedural autonomy, run opposite to self-government and frustrate the exercise of autonomy, and hence should not be accepted as autonomous. On this reading, autonomy is a value with a definite content, namely the capacity to determine one's life in a way that is consistent with the assertion of one's will. ²² Not all ways of life are instantiations of such a capacity, and those that forfeit self-direction through submissive choices count as non-autonomous. Such choices may correspond to the agent's wants and might even realize her well-being under certain circumstances, but autonomy should not be confused with liberty, on the one hand, and well-being, on the other. ²³

The substantive view cashes out its appeal from the difficulties of the procedural model with reference to the Stoic implication ²⁴ and to the choice of enslaving oneself, ²⁵ but it exhibits its own shortcomings. Consider first the Stoic implication. According to the procedural model, so the objection goes, if the agent is a Stoic and reshapes her expectations so as to free her desires from external constraints, she does not suffer from an autonomy impairment, as she is indifferent to external contingency, no matter how limited her life circumstances. Now, as it happens, the Stoic implication has been addressed in a procedural fashion by means of the distinction between adaptive versus autonomous preferences. ²⁶ However, the distinction cannot differentiate preferences which were originally “mine” from those which were adaptive from the start, being vulnerable to the same two problems faced by procedural autonomy, namely infinite regress and incompleteness. Substantive views, in this respect, seem to sidestep the authenticity issue by defining normative constraints on options and preferences independently of their origin. ²⁷ Under such light, the Stoics' retreat in the inner citadel does not count as autonomous. But here we also reach the crucial limit of substantive views of autonomy: if independent normative criteria select non-autonomous from autonomous choices, one may question how any choice can be called autonomous if it is not recognized as such by the agent. In order to be a coherent ideal, autonomy requires at least minimal internalism.

Consider now the objection of self-enslavement. How are we to accept the choice to enslave oneself, even where it is reached correctly according to the procedural model? Mill's rejection of the liberty to enslave oneself is argued on the basis that one cannot be free to make oneself unfree forever. Thus, contrary to what supporters of substantive autonomy seem to imply, it is the lack of exit that makes such a contract normatively invalid for Mill, not the evidence of lack of autonomy in the content of the choice itself. We can in any case concede that the choice to enslave oneself, with its implication of forfeiting one's future autonomy altogether, can hardly count as autonomous. If, however, our intuitions about self-enslavement are extended to analogous though less extreme cases, as suggested by the supporters of the substantive model, the stretching may prove too much and the argument against procedural autonomy may consequently fail. The example of self-enslavement is used to ground the position that choices, no matter how procedurally autonomous, can pass the test of autonomy only if they concern ways of life which do not give up self-government. Thus not only self-enslavement, but also, for example, the choice to submit to one's family and to patriarchal rule or the choice to enter a monastery and become a nun, even if produced by a critical deliberation and a wholehearted endorsement of their underlying values, is considered non-autonomous, for they imply the obedience to rules and authority other than the self. But if it were true that one cannot autonomously accept an external authority, for it would imply giving up self-government, then, similarly, the anarchist challenge to political authority would stand and tear the political obligation to any government to pieces. ²⁸ When we move from self-enslavement to less dramatic examples, the intuitions supporting substantive views of autonomy are actually challenged. Moreover, so far the substantive views have not yet provided a satisfactory response to the question of how a life can be defined as autonomous without it being chosen autonomously.

Autonomy and public reason

Which model of autonomy can better ground and justify a lack-of-autonomy judgment? It would seem that a substantive conception of autonomy would more directly lead to such judgments: if a life of submission has been declared incompatible with autonomy, an external observer need not know anything more about the agent's history and preference structure to judge her life as autonomy lacking. From the point of view of strong substantive autonomy, then lack-of-autonomy judgments can be drawn and would be the preliminary condition for an autonomy-based argument stressing the harmful consequences of a certain conduct for the self, hence calling for limiting toleration concerning a particular practice. If such an argument withstood scrutiny, then interference with that practice could in principle be justified. However, this very judgment would be highly controversial even from the point of view of “weak-substantive” autonomy, ²⁹ and, of course, would be rejected by all proceduralists.

Now liberalism of all varieties implies the fundamental principle that “all interferences with action stand in need of a justification.” ³⁰ Moreover, the justification should be acceptable to all citizens, including the people whose liberty is interfered with. In other words, justification should be framed in terms of public reason. But, then, how can a contentious view of autonomy, leading to an even more contentious judgment of lack of autonomy, constitute the appropriate basis for a public justification of interference, in principle acceptable to the very people whose liberty is restricted? Clearly it cannot. Autonomy-based arguments for questioning toleration of certain practices are therefore poor candidates for justification by public reason.

Autonomy and liberalism

Thus far I have critiqued autonomy-based arguments for justifying political intervention in people's life given the controversial nature of their underlying conceptions of autonomy. From this criticism it follows that political action aimed at promoting autonomy concerning people's life and choices judged as non-autonomous is not justified. Yet, liberal democratic politics is certainly committed to the ideal of autonomy, which is indeed integral to liberal-democratic legitimacy in a crucial sense. ³¹ So we need to ask how come autonomy is a crucial liberal value which however cannot ground political action aimed at its promotion. More precisely we must understand which conception of autonomy is crucial for liberalism, what is its role in the construction of liberal legitimacy, and whether it is closer to the procedural or the substantive views or, indeed, whether it is something altogether different.

To put it briefly, autonomy intersects liberal theory in the justification of liberal institutions, given that in order to be legitimate political principles and fundamental rules ought to be endorsed by whom they will apply. The endorsement condition assumes that citizens are considered able to make autonomous choices. Then autonomy intersects democracy for democratic rule assumes that citizens are co-authors of the law hence capable of collective self-rule. This double political role of autonomy, in liberal and democratic theory, has been interpreted by John Christman by the definition of political autonomy as a “status marker of citizenship.” ³² He says that autonomy is attributed to all citizens as a condition for being, and being considered, a citizen; such attribution is grounded on the presumption of the presence of basic autonomy in all adults without special cognitive and mental disabilities, and freed from oppressive and constraining conditions. ³³ Basic autonomy consists in a minimal status of being responsible, independent, and able to speak for oneself, that is to say in basic agential capacities, which in turn are the basis for equal respect. Christman's suggestion is to my mind crucial for outlining a political conception of autonomy; his interpretation, however, does not press further the status marker suggestion, but rather attempts to establish links between this political conception and the two conceptions of personal autonomy. ³⁴ More precisely, he asks which idea of personal autonomy is implied in the political conception, and establishes a parallel between the dichotomy of the two main conceptions of autonomy, procedural and substantive, and the political dichotomy of neutralist and perfectionist liberalisms. In his view the procedural view, being content-free and value-neutral, fits the framework of liberal neutrality, while the substantive view of autonomy embodies a perfectionist conception of a person that sits well with liberal perfectionism. My first critical remark on Christman's view is that the procedural model, insofar as it is not taken to be purely descriptive, is not value-neutral. After all, it entails the (procedural) values of autonomy, and procedural values are still values. The values in question here are precisely those of internal critical reflection on one's preferences, commitments, and dispositions to act in certain ways; they are the values of building one's character over casual, traditional, and habitual behavior, and of the examined over the unexamined life. “Procedural” does not mean “content-free” but rather “open-ended,” and the autonomous value of what is chosen is provided by the process by which it is brought about. Then the content of the (autonomous) choice has an independent value for the agent as well, in so far as it fulfills her well-being, her satisfaction, her commitments, and so on. ³⁵ Whatever the value of the choice content, the value of the autonomous deliberation is a different and independent thing. The procedural value of autonomy is indeed compatible with many values at the level of choice content, including the affirmation of one's autonomous will and the promotion of one's future autonomy. Again, this much does not imply that procedural autonomy is value-neutral. Nor does it imply value relativism as maintained by supporters of the substantive view of autonomy. In fact, like the substantive view, the procedural value of autonomy may subscribe to a perfectionist view of the person; they differ, in this respect, only in how the value of autonomy is instantiated: via critical reasoning and procedural independence in the case of the proceduralist account, or via certain outcomes of choice which are taken to underwrite the autonomy of choice in the case of substantivists. So the liberal perfectionist is, I think, able to endorse either conception, in principle.

For the same reason, both conceptions look unfit to be part of neutral liberalism, and this is my second remark on Christman's parallel. ³⁶ Both views of autonomy are contentious and controversial, and thus are unsuitable as the political principles needed for building liberal legitimacy that aims at being independent of any comprehensive view or value, and for the same token neither can be part of public reason grounding legitimate political action. Here, I have in mind the kind of political liberalism spelled out by Rawls ³⁷ and others after him, ³⁸ which constitutes a special and, to my mind, more convincing interpretation of neutralist liberalism.

The ambition of a political justification of liberalism is to be persuasive to people across (reasonable) pluralism and possibly beyond those who are already convinced liberals. Hence, it attempts to provide a liberal justification, which, by being independent of conflicting comprehensive views and appealing only to political principles, claims to be freestanding and acceptable to all. What is needed to build such a political conception is the idea of citizens as free and equal persons which clearly is not a neutral standpoint, but which, in its political dimension, is taken to be a plausible premise shared across contemporary reasonable pluralism. The two conceptions of autonomy previously considered are definitely controversial, belong to comprehensive views of interpersonal relations and ethical life, and are inadequate as political ideals and grounds of liberal political morality. Below, I will demonstrate their political inadequacy in our democracy with the help of some examples. But now I would like to pursue the conception of autonomy that is relevant to political liberalism, the one which has actually been suggested by Christman, but needs further reflection.

Autonomy as status marker

Some liberal thinkers have stressed that what is needed for liberalism is moral autonomy. ³⁹ While personal autonomy “evokes the image of a person in charge of his life, not just following his desires, but choosing which of his desire to follow,” ⁴⁰ moral autonomy, of Kantian provenance, is the capacity to be moved by moral reasons. The two ideals are distinct, but not independent, given that the capacity required by personal autonomy—self-reflection and critical distancing from one's first-order motivations—is the same as those required by moral autonomy; a person cannot be capable of moral autonomy if she is not able to be a self-directed agent. ⁴¹ This is what Gaus calls the “ultra minimal view of personal autonomy,” which, to his mind, is all that political liberalism requires. ⁴² Gaus' view resembles Christman's basic autonomy, but I think that what is relevant for political autonomy is not how thick or minimal the conception of personal autonomy should be for political liberalism to stand, but rather the distinctive role played by autonomy, both as a personal and as a moral capacity, in the building of liberal-democratic legitimacy, which is radically different from that played in moral and social life. As said, autonomy constitutes a status marker of citizens, and, in this sense, is the premise for a democracy to work as a form of self-government and to be justified as a political order. As a status marker, autonomy is ascribed to all citizens on the basis of the presumed capacities of moral reasoning, independence of judgment, and of being able to speak for oneself and be responsible for one's actions. Being ascribed in this fashion, the status of autonomous person is unconditional; it is not dependent on the actual capacities for autonomy differentially displayed by individuals, and so the status does not come in degrees. This, then, is the distinctiveness of the political conception of autonomy: neither a capacity that is empirically displayed by individuals in their characters and actions, nor a metaphysical property of the rational will, it is instead simply a reasonable assumption required to conceive of persons as free and equal.

Yet how can we grant the reasonableness of the assumption given all the empirical evidence of the enormous variations among and within individuals, of the capacity for autonomy? One possible answer is the Kantian demarcation between empirical individuals and noumenical persons. This solution, however, is not available to political liberalism as the latter cannot rely on Kantian metaphysics, which is a contentious philosophical view, for outlining its putatively freestanding political conception. An alternative way of accounting for the empirical variations in autonomous agency without giving up the unconditional ascription of autonomy is to understand the capacity for autonomy as a “range property.” A range property is a nonvariable property (it either exists or does not exist) that supervenes on a particular range of variation of a variable property. ⁴³ We can indeed reasonably take it that all persons possess an autonomous capacity above a minimal threshold and that is all that is required for granting the ascription of autonomy as a status marker. The minimal threshold is precisely the basic autonomy described by Christman which is assumed to be present in all adults without mental and cognitive disabilities and in condition of freedom. As it works, in order to be ascribed the status of autonomous citizen, no one needs to pass the test of the minimal threshold, but it is rather true the opposite: only when someone manifestly shows to lack such basic capacities, after medical inquiries, is declared unfit to plead.

The range property solution is hinted at by Rawls when he states that the presence of the two moral powers (the capacity for a sense of justice and the capacity for a conception of the good) characterizes persons and that variations within a range from a minimum to a maximum level do not affect the characterizing property. ⁴⁴ Individuals must be assumed to be autonomous persons both for the justification of liberal justice and for the working of a stable and just political order. This assumption constitutes the background premise of the whole argument. It is necessary not only for choosing the principles of justice and enabling citizens to reason from, act upon, and be ruled by those principles, but also as the grounds for equal respect. Autonomy underwrites human dignity, which commands equal respect.

In developing the complex architecture of the political conception, more specific concepts of autonomy are introduced as crucial for the justification and the stability of the well-ordered society. As far as the justification is concerned, the constructivist argument of the original position implies seeing the parties as rationally autonomous. Rational autonomy is indeed a form of procedural autonomy, but, as stressed by Rawls, it is an “artificial” feature, a device of representation for modeling the original position. In other words, in order to provide a representation of the original position, the parties must be endowed with rational autonomy or “the capacity to form, to revise, and to pursue a conception of the good, and to deliberate in accordance to it.” ⁴⁵ Finally, in the public life of a well-ordered society, citizens realize full autonomy “when they act from principles of justice which specify the fair terms of cooperation they would give to themselves when fairly represented as free and equal persons.” ⁴⁶ Put simply, citizens are fully autonomous when they act as citizens on the basis of public reason. Rational autonomy, as I have said, is a form of procedural autonomy; in contrast, full autonomy is a strictly political concept specifying the proper kinds of reasons that ought to motivate autonomous citizens. And this sounds like a form of substantive autonomy. But these more specific concepts of autonomy have a limited role and a precise function at a certain stage of the argument that does not add or subtract from the crucial role of the general concept of autonomy as a status marker.

In sum, political liberalism presumes autonomy as a rational and political capacity ascribed to all adults without mental and cognitive handicaps, and this presumption confers the status of citizen, as locus of equal respect and rights and co-author of the law. But it does not require that citizens actually display personal autonomy (or moral autonomy, for that matter) in their daily lives. While autonomy is likely to figure in the moral outlook of many liberal citizens and in a prominent way, this much does not qualify autonomy as something to be promoted politically by means of policies aimed at restricting supposedly non-autonomous choices and actions. This view is reflected in the democratic practice to grant equal political rights to all adults, without mental and cognitive disabilities, without testing their level of personal autonomy. Finally the view of autonomy as a status marker, based on a range property, does not make the concept redundant in politics. The focus on the agential capacity in liberal-democratic politics as the proper status of citizenship invites considerations of justice criticizing the persisting unequal conditions for practicing autonomy and advocating fair distribution of equal opportunity for being or becoming autonomous.

The problematic political implications of autonomy

My argument for a purely political conception of autonomy for the purposes of liberal-democratic politics counters a conventional position, according to which the conceptual analysis of personal autonomy is taken to have political implications, especially concerning the possibility of justifying paternalism, under specified circumstances. ⁴⁷ But drawing political implications from the analysis of personal autonomy can have dangerous political consequences for the core morality of liberal democracy. In the first part of my discussion I tried to show that toleration cannot be limited by lack-of-autonomy judgments, which are difficult to make and justify because of their controversial grounds. Here, however, I want to advance a more general argument against making political use of comprehensive conceptions of autonomy.

Political implications are usually drawn because the value on autonomy is seen as the crucial barrier against unchecked paternalism. ⁴⁸ If the reason why paternalism is generally unjustified lies in respect for autonomy, then, conversely, and as many scholars have argued, lack of autonomy may justify paternalistic intervention since there is no autonomy to respect. ⁴⁹ Lack of autonomy does not alone provide the justification for any paternalistic intervention, but it is usually coupled with arguments for self-harm either of the individuals or even of the whole groups of which the individual is a member.⁵⁰ On the whole, the argument for paternalistic intervention crucially depends on the possibility of reaching a lack-of-autonomy judgment about individuals' choices and actions. We have seen earlier that such judgments are not easily made. But they can in principle be made, which is why I wish now to focus on the consequences of this line of argument for paternalistic intervention.

First, lack-of-autonomy judgments can in principle be made, especially, though not exclusively, on a substantive view of autonomy. Although few scholars maintain that lack of autonomy is a sufficient condition for paternalistic intervention, it is clear that, from the point of view of supporters of strong substantive autonomy, the promotion of future autonomy, coupled with the prevention of self-harm, is a reason for intervening in choices that jeopardize autonomy, such as a life of submission. Second, and in contrast, advocates of the procedural model typically oppose paternalism. Certainly they do not share the view that the outcomes of choices can ever constitute a good reason for intervention, since, for them, autonomy is defined by procedures and not by outcomes. As noted previously, it is thus much more difficult to test procedural autonomy in real cases as individuals' mental states and processes of deliberation are not readily accessible to onlookers. Yet, in principle, even the procedural model can be applied and test an agent's autonomy. For example, certain judgments are made easier by the fact that some conditions are more verifiable than others, such as adaptive preferences or a lack of cognitive lucidity. So both the substantive and the procedural models can be used to test people's actual autonomy, at least theoretically.

The problem is that testing citizens' autonomy implies making judgments on the basis of drawing distinctions between autonomous (more deserving?) and less autonomous (less deserving?) people. Deserving of what? I would say “respect,” if respect is grounded on the basis of agential capacity. In this case, we would have shifted from autonomy and equal respect as status markers of equal citizenship, ascribed apriori to all, in virtue of the range property, to autonomy by degree, akin to what Stephen Darwall ⁵¹ calls merit- or esteem-respect that is attributed aposteriori, on the basis of desert. If political theory invokes autonomy not simply as a status marker or on a strictly political understanding, it risks drawing (moral) distinctions among: (a) fully autonomous citizens—those Kantian or Millian agents—who deserve to have their actual autonomy protected, (b) citizens whose autonomy is under probation, and (c) those who are clearly not autonomous, hence are more minors than equals, and who warrant paternalistic intervention so as to foster their potential future autonomy. I think such distinctions among citizens are unacceptable and not just from the point of view of political liberalism, but of liberal democracy in general. Moreover, distinguishing among citizens who are fully, only partially, and decidedly non-autonomous tends not to be applied consistently, but rather channels social and cultural differences. In so doing, it produces conspicuous cases of double standards, as we shall now see.

The practice of autonomy

How the value of autonomy is understood in politics has serious ramifications for toleration, as I noted at the outset. Toleration means the suspension of the power of interference with actions, practices, and behavior to which the potential tolerator has reasons to object. Given that it is not immediately clear why allowing something that we disapprove is a good thing, for toleration to be a virtue, its justification is crucial.

A standard justification is held to be respect for people's autonomous choices. Thus, if autonomy is meant not as an ascriptive capacity shared by all persons or “status marker,” but rather as a comprehensive moral ideal or critical capacity, then the autonomy behind the choice in question may not be taken for granted. Testing the choice according to either the substantive or procedural models considered earlier may reveal it to be non-autonomous. In that case, tolerating the choice would be unfounded; there would seem to be no reason for potential tolerators to suspend their inclination to interfere with a choice of which they disapprove. Interference with others' behavior would simply appear the right thing to do. But in this case, it is clear that the virtue of toleration would be too limited, the tolerator too self-righteous, and that too much interference in other people's choices and behavior would be licensed.⁵²

Take the example of the hijab. Something like the understanding of toleration above, informed by a comprehensive concept of autonomy, often underpins arguments against tolerating the hijab. A major objection to its toleration is, for instance, that it is a symbol of women's submission to men. The apparently free choices of the girls and women who wear the hijab are discounted for not being autonomous, being regarded, instead, as the result of oppression, and so not warranting toleration. ⁵³ To be sure, there are other arguments for prohibiting the hijab, trespassing on the ideal of laicité being the main one in France. But the issue of the autonomy of women has been crucial throughout the debate and has helped to ground the prohibition of the hijab in state schools even in France.

And yet, instructively, autonomy testing has never been applied to majority practices and behaviors such as piercing, extreme dieting, tattooing, plastic surgery to name a few. Such practices, which are just as dubious from the standpoint of substantive autonomy, have never constituted issues for questioning toleration in liberal democracies. It seems that the only relevant difference between them and hijab wearing is that the former are familiar practices of our culture; though they may well signal the absence of autonomy, they are simply taken for granted. The women who choose piercing and plastic surgery are not tested as to their autonomy; no question is asked about whether they deserve autonomy protection or whether they are harming themselves, hence need protection from their lack of autonomy.

I am not saying that piercing and plastic surgery should as well be considered as displaying lack of autonomy and leading to self-harm, hence justifying paternalistic intervention, for I do not hold lack of autonomy and self-harm as valid justification for limiting toleration. I am just stressing the double standards at work in this respect, when at issue there are “our” practices versus “stranger” practices. The fact is that for “our” practices, no matter how questionable from the point of view of substantive autonomy, no one has ever advocated a public ban, while such ban, for example, has been at the center of the long controversy which led to the Stasi regulation in France prohibiting the headscarves in school. True, the regulation establishes that no ostentatious and provocative religious symbol should be worn at public offices, but it is a fact that until headscarves had appeared in state school, no one has ever complained about kippah or cross wearing. The laicité argument has been clearly contaminated by the consideration of hijabs as symbols of women submission, as the wide public discussion has amply documented. Thus in the attitude toward practices that may look dubious from the point of view of substantive autonomy, on the one hand, there is a presumption that all preferences and options in our culture are autonomous by definition, and, on the other, that those linked to other cultures are instead suspect. Why should we want only women of alien cultures to conform to high standards of personal autonomy? I think that here we have touched a clear case of double standards, and of biased judgment in favor of what is familiar against what is foreign, strange, and thus suspicious and seemingly dangerous. True, the double standards problem may be solved by applying the same stringent test for autonomy also to our mainstream practices, but this would lead to excessive state interference into personal freedom, which could not be justified in public reason.

Now, the problems of dubious distinctions among citizens and of double standards do not arise if the autonomy and respect figuring in the justification of toleration are taken as status markers. On this account, fellow citizens owe respect to each other as equals and are presumed equally capable of autonomous choices, much as we expect others, from their different moral standpoint, to respect “us” and “our” autonomy. Autonomy and respect as status markers thus confer reciprocity on toleration, which would otherwise remain entangled in an asymmetrical power relation.

Some possible objections

It might be objected that my conception of toleration, based on autonomy only as a status marker, lacks the discriminative capacity to discern the intolerable. Toleration, after all, can be a moral good only within limits. Does not only the hijab, but also the burka, female mutilation, forced marriage, and the whole pack of patriarchal oppressive practices affecting women end up being tolerated on my easy-going, status-marker view?

The answer is that limits of toleration are also acknowledged in my approach, but such limits are fixed by a reasoned interpretation of the harm principle. What counts as harm is contested, and that is why the interpretation of harm should be as widely shared as possible. I do not want here to discuss the harm principle, but rather stresses that, on the basis of the harm principle, limits to tolerance and to justified interference and paternalism are actually in line with the principle of equal respect and autonomy as a status marker. Here, a judgment that intolerable harm is being perpetrated does not depend on a judgment of people as minors, much less that they are incapable of self-directing their life, and hence are less than full citizens.

If respect for the equal status of citizenship is what prompts a preference for the harm principle, this fact then constrains the possible interpretations of harm relevant to establish grounds for intervention. Self-harm, for example, constitutes an open issue for a political perspective that ascribes autonomy unconditionally to all. Such a perspective seems to imply that no interference with people's choice is permitted, out of respect of their (ascribed) autonomy, unless harm to others is involved; hence it seems to dispense with looking into actual choices leading to self-harming outcomes. Respect-based considerations may still ground paternalistic interventions, but only in the two following cases.

First, there is a class of potentially self-harming action that has a damaging effect on the society as a whole. This is the case with seat belts and motorcycle helmets where paternalistic intervention is justified by the prevention of harm to society as a whole. In such cases, there is no need to consider the autonomy of actions, because we have independent grounds for intervention.

Second, the unconditional ascription of autonomy as the range property making us all equal citizens is not incompatible with the acknowledgment that such capacity is actually displayed in a wide range of variations, even in the same individuals under different circumstances. The actual intrapersonal variations do not threaten the possession of the range property should they fall below a given threshold, and hence do not call into question the ascription of equal citizenship status. As a rule, equal respect for the autonomy of citizens requires that their choices and actions are not scrutinized too closely and that we credit their capacity for choice. ⁵⁴ However, when there is clear evidence that certain choices will bring about serious and irreversible bodily harm, then there is a good reason to consider the nature of the choice and whether the internal and external conditions that make it normatively valid are present: whether the agent's faculties are impaired and whether she is free from the pressure of threats and manipulation.

If it turns out that internal or external conditions for a normatively valid choice are not present, then there are grounds for preventing or stopping the self-harm and paternalistic intervention is justified in such cases. ⁵⁵ But again, there should first be evidence of serious and permanent bodily harm as a consequence of a certain choice; it is this evidence that authorizes political institutions to look into and behind the choice, to make sure that the conditions for normative validity are present. In the absence of evidence of bodily harm of a serious and permanent kind, no scrutiny is authorized; symbolic self-harm, for example, cannot set the process of scrutiny in motion. It also follows that serious and permanent bodily mutilations are only a necessary but not sufficient condition for paternalistic intervention, as the case of sex change shows. As I have argued elsewhere, the serious and permanent bodily mutilation involved in sex changes is typically chosen under conditions granting normative validity, hence no paternalistic intervention is generally justified.⁵⁶

In sum, paternalistic intervention is generally ruled out by the principle of autonomy as a status marker of citizenship. In special circumstances, however, paternalism can be justified, and that is the case when (a) prospective self-harm causes damage to the society as a whole, and (b) when self-harm consists in serious and permanent bodily damage for the person; this fact authorizes institutions to look into the agent's choice and see whether the conditions for normative validity hold. If they do not, paternalistic intervention may be justified.

Conclusion

The main point of contention between the procedural and the substantive conceptions of autonomy lies precisely in their implications for the justification of political interference with certain practices. I have argued that political interference with people's choices cannot follow from the lack of autonomy on any formulation for two different moral reasons: First, a principled reason, concerning the fundamental commitment to the equal status of citizens, and, second, a reason of justice, concerning the double standards that are usually displayed by paternalistic interventions for the promotion of autonomy.

My argument might be taken to mandate an “anything-goes” attitude to any cultural practices, even those that are suspect in relation to the rights of individuals in the community, or which are obstacles to cultural integration and the social order. However, this is not what is implied by my argument. I have argued for a political definition of autonomy and against a comprehensive notion of autonomy in the area of democratic politics, having tried to show the problematic implications of making use of a comprehensive notion for the theory and practice of liberal democracy. As far as theory is concerned, comprehensive autonomy may collide with the principles of equal citizenship and equal respect for persons. Concerning democratic practice, it may license unjust forms of paternalism and double standards for different sets of citizens.

Giving up a comprehensive conception of autonomy though does not mean being indifferent to the oppressive situation of women in certain cultural communities. ⁵⁷ What it does means is that political intervention in this respect should be oriented to the protection of rights and the prevention of bodily harm, and not to the fostering of women's autonomy.

I acknowledge the special difficulty of political action for the protection of members of some cultural minorities, who are caught between liberal bias against non-liberal culture and liberal indulgence of cultural respect. And with reference to autonomy, understood as the actual fulfillment of the capacities with which each person is endowed, which is the concern of this essay, democratic politics has indeed a responsibility, and that is granting fair equal opportunity for its exercise. Equal opportunity for autonomy implies both distributive measures—such as education, jobs, and income—and symbolic measures concerning recognition of minority members as equals given their identity and differences. Both measures work to sustain equal citizenship and equal respect and to provide the social basis of self-respect. It would then be desirable that social work could contribute in this direction, providing support and alternative options to traditional practices; and this is the task of civil associations and society that can definitely criticize oppression of all kinds and try to persuade victims to opt out oppressive institutions, providing them with actual help to sustain them. Beyond that, no shortcuts are available, because autonomy, like faith, cannot be imposed if it is to retain its value.