1. Introduction

We refer to autonomy as that human capacity to form a personal opinion and acting according to it, where the confrontation with information deriving from different contexts is fundamental for developing it.The contemporary debate shows that most of the authors tend to overcome the metaphysical question of determinism and free will because they are oriented toward the concrete dimensions of individual choices and the role of socialization to develop capacities for critical reflection.  The debate on autonomy is very lively in different fields. There are therefore many conceptual distinctions worthy of consideration. For the sake of my discussion, the most important distinction is between "moral" autonomy and "personal" autonomy. Personal autonomy means that we consider autonomy not only in a strictly moral sense (like the Kantian perspectives).

We can describe personal autonomy in "procedural" or in "substantive" terms. Procedural theories emphasize the structural conditions of the process of "identification" with one's own motives. Even if these conditions are relevant, substantive theories rightly point on the role of the content of our reasons for autonomous agency. This perspective requires substantive standard according to which we can recognize and criticize oppressive norms.In our opinion, a person is autonomous if she is able to recognize and to take responsibility for subjective reasons (i.e. desires and preferences) and objective reasons (moral norms) for acting. Moreover, if we adopt a pragmatic model, it emerges the know-how implied by the "social role" of the autonomous agent as "scorekeeper", who participates in the game of giving and asking for reasons. This social role is defined by the use of language bound to certain social attitudes (attributing and undertaking commitments and entitlements) through which the recognition of deontic statuses (commitments and entitlements) seems possible. Even if we accept the inferential structure of this space proposed by Wilfrid Sellars, we must also give an explanation of the social perspectives from which we can undertake and attribute commitments. Starting from this thesis, we can conclude that the autonomous agent occupies the social role of scorekeeper, thus she is able to justify and to take responsibility for her assertions (or the assertions of others). The acknowledgment of reasons is structured by the deontic structure of discursive practices. In this sense, we as agents are in the role of scorekeeper when we are able to attribute reasons for acting and to undertake the corresponding commitments by ourselves if it is the case. This happens in the interaction with other people, namely when we get in touch with points of view different from our own. This is the reason why we attribute a fundamental role also to informational and social diversity.

2. Substantial Autonomy vs Procedural Autonomy

Procedural personal autonomy requires the fulfillment of conditions for rationally deciding and acting, ranging over a wide spectrum of individual idiosyncratic desires and volitions (Dworkin, Frankfurt, Ekstrom). Gerald Dworkin and Harry Frankfurt introduced the higher-order desire approach to present the structural condition of autonomy as "authenticity". The reflexivity essential to self-rule can be explained in terms of the attitudinal relationships between the person's wants, or what are called optative relations. Dworkin distinguishes between liberty and autonomy. We need a separate notion of autonomy because not every interference with the voluntary character of one's action interferes with a person's ability to choose his mode of life. The condition of procedural independence introduces therefore an infinite regress: if the acts of identification must themselves be autonomous, there must exist another act of identification at a higher level. A first solution to this problem could be the claim that desires can be autonomous without foundations. But this option makes it impossible to recognize whether an action is the result of manipulation. The problem is to explain why these conative attitudes have the authority to ground the agent's standpoint. The reason resides in their nature: they are plan-type attitudes and not mere desires. Our planning agency extends over time; this extension involves activities at different times performed by the very same agent. Thus, plan-type attitudes form the temporally persisting agent. Why are they higher-order attitudes rather than first-order ones? According to Bratman, plan-type attitudes are higher-order attitudes as they show the agent's self-governance by appealing to considerations that legitimize and justify autonomous choices. What is very important here is the shift from the motivational to the "normative" content: the content of second –order attitudes is not desires but "self-governing policies" motivationally effective in practical reasoning. Self-governing policies ground the agential authority through their "reflexivity", that shows its efficacy in case the agent endorses the content of such activities i.e. she is satisfied (in the sense of Locke) with them.

Several authors add "historical" conditions which examine the process of the formation of judgment or decision to act (Fisher, Ravizza, Christman). According to Christman, It is a matter of the way in which the desire was formed i.e. the conditions and factors that were relevant during the "process" of coming to have the value or desire. It seems easier to recognize whether that actual desire of the agent is authentic, especially in the case of formation of a totally new desire, if we consider the conditions that determine the agent's participation in the process of preference formation. Before setting out these conditions, a fundamental requirement must be satisfied: that the agent had the possibility of resisting the development of a desire and she did not. An autonomous person must show not only "procedural" (process or conditions for deliberation) but also "substantial" independence, which rightly requires the consideration of the "social context" in which an action can be judged as autonomous . Very interesting and powerful are the substantive theories, that in the individual variant start from the Kantian inheritance (Korsgaard, Hills, Wolf). The debate on autonomy presents several "substantive" perspectives (Wolf, Benson, Oshana, Mayer, Stolyar among others) that – generally speaking – try to overcome the "regress problem" involved by the procedural theories (both in their structural or historical variants).

Following Meyer's criticism to the "unified self", Benson notices that identity-based theories set conditions too strong to be necessary for autonomy, because one can take ownership for what he/she does even if the action does not align with who he/she is or what he/she stands for. A fruitful example of this situation is the performance of trivial acts such as ‘swivelling my office chair" that rise above the level of sub-intentional behaviors. These acts whenever subjected to critical scrutiny could generate alienation from the ways in which the agent was moved to do such trivial things. Nevertheless, the agent is autonomous in performing them. But trivial acts are problematic for procedural identity-based theories because they directly challenge the relationship these theories presume between what the agents care about and which actions are authentic. Another deeply problematic situation is the possibility of integrating different aspects of the self. For an autonomous agent it becomes very difficult to take ownership for commitments that are incompatible even if they constitute the identity of a single person and so they generate internal conflicts. Identity based theories cannot explain autonomy in such cases because they require "identification" with a motive as necessary condition for reflection.

3. Social Conceptions of Autonomy

Habermas' account of "communicative action" represents a good social model which relates autonomy to a linguistic normative competence, and even though it is intersubjective, still it is procedural. According to Habermas, autonomy is bound to the acknowledgment of presuppositions or linguistic rules as conditions of universal validity of theoretical and practical claims.

Differently from other theories of "recognition" (Honneth, Taylor), he introduced the issue of "interpersonal recognition" related to the formal linguistic conditions of a rational and egalitarian dialog (communicative action). The development of the capacity for autonomy through communicative action is based on a well-known process of socialization explained also by reference to different disciplines (Piaget, Kohlberg). In this context, Habermas establishes a fundamental relationship between autonomy and "communicative action". He interpreted the Meadian concept of identity in a pragmatic sense. Mead maintains that the formation of the identity develops through the medium of linguistic communication. The process of socialization is a process of individualization based on an asymmetry between the perspectives of speaker and listener. The "self" is the identity of the socialized individual who has undertaken fundamental roles in a linguistic situation. The self indicates the point of view that Ego presents to Alter in the interaction, when the latter makes an offer of speech act. The interpersonal relationship between speaker and listener is fundamental as Ego, by undertaking the perspective of the interlocutor, cannot abandon his/her communicative role. In this sense, Ego undertakes the perspective of Alter for picking up his/her expectations; Ego is the first person role that must satisfy the behavioral models at first undertaken and internalized by Alter. According to Habermas, the performative attitude assumed by Ego and Alter in the communicative situation is bound to the presupposition that the interlocutor has the possibility of accepting or refusing the offer of speech act. Ego cannot give up this "space of freedom" even in the case of the playing of the social roles; indeed the very internalized behavioral model implies the linguistic structure of the relationship between "responsible" (i.e. autonomous) agents.

In our opinion, Habermas does not provide a good explanation of the "space of freedom" of Ego. He intends autonomy of the agent in pure procedural terms without considering the role of the normative structure of the semantic content on the objectivity of linguistic validity claims. According to Habermas, an agent is autonomous only if she has the consent as her fundamental end. For an agent to reach this end she must satisfy some structural conditions of the intersubjective linguistic practice: (a) pursue without mental constrains their illocutive ends, (b) subordinate their consent to the acknowledgment of validity claims and (c) be available to undertake commitments, that influence the development of the interaction.

Procedural theories underestimate the role of the internalization of oppressive norms; from a substantive point of view, even if a person has to some extent the option of choosing alternative values, it is the "content" of the norms she internalized that diminishes her autonomy. We must consider the "nature" of values that come from the social context. For instance, Benson focuses on a fundamental point in order for understanding the normative source of autonomy. He addresses directly to the social and discursive dimension of "taking ownership" that explains how "internalized invisibility" (internalization of oppressive norms diminishing autonomy) can defeat agents capacity "to take ownership" of what they do. The active dimension of taking ownership implies the capacity of the agent of giving reasons for her actions and so of responding to potential "challenges" arising in the social context from her own point of view.

The active dimension of taking ownership implies the capacity of the agent of giving reasons for her actions and so of responding to potential "challenges" arising in the social context from her own point of view. The active character of ownership can be clarified in three central points:

(1) Claiming authority for ourselves as ones who are in a position to speak for our conduct is not a matter of deliberate action;
(2) Self-authorization arises partly out of our self-regard and it transpires within the reach of our capabilities to reflect, decide and act.
(3) Taking ownership of our actions is also a matter of taking responsibility and this active dimension could not be the result of deliberation.

We maintain that Benson's account rightly points to the social and discursive dimension of autonomy. This move gives the possibility of taking responsibility in a public context and implies also the possibility of speaking for people who are marginalized ^[1].

4. Autonomy as a social role

In the paper Autonomy, Community and Freedom, Brandom shows a tension between attitude-dependence and content-dependence when we consider autonomy and freedom. Brandon tries to reconcile Kant and Hegel classical opposition. Following Kant, we are genuinely normatively constrained only by the rule we adopt and acknowledge as binding on us. He moves from this thesis the consequent idea that the capacity to be bound by norms and the capacity to bind ourselves by norms are one and the same. Authority and responsibility are symmetrical and reciprocal because they are constitutive features of the normative subject who is at once authoritative and responsible. But, this move is possible not in virtue of the Kantian notion of autonomy that applies in the moral field and refers to the substantive principle of the categorical imperative but in virtue of the tension Brandom establishes between attitude-dependence and content-dependence

Brandom seems to collapse autonomy into positive freedom because he invokes Hegel when he underscores the nature and the content fulness of concepts. Hegel stresses on the fact that contents must have a kind of attitude-independence, consequently the content acquires a sort of authority that is independent of the responsibility that the agent takes for it. The tension between attitude-dependence and content-dependence is solved with the introduction of the social model of reciprocal recognition: authority and responsibility are ultimately social phenomena. Reciprocal recognition becomes dependent from the fact that we necessarily refer to joint commitments which have material inferential contents. This result entails that freedom and autonomy have the same meaning: agents are autonomous or free because they necessarily bind themselves to shareable commitments i.e. commitments accepted by their community. According to this result is the game of giving and asking for reason a game that agents play only into the boundaries of their community? What about the possibility of finding a universal pragmatic structure that favors the dialog among cultures?

Autonomy is an essential component of the self-realization of a subject living in a society that develops in communication, i.e. in the intersubjective acknowledgement of commitments or validity claims. Autonomy can be considered as that capacity human beings normally have of decentralizing their own point of view, thus of distinguishing subjective and objective reasons. It is guided by conceptual rules and for this reason only requires an analysis of the structure of the process of their recognition. Autonomy doesn't mean rational choice (means-end reasoning), but capacity of participation to the "social", hence discursive "game of giving and asking for reasons". Autonomy can be thought in Kantian terms as acting according to our conceptions of rules. This "normative compulsion", which is quite different from the natural compulsion, forces us to act according to our "grasp" or "understanding" of rules. The compulsion of rules is mediated by our attitude toward them, i. e. we must "acknowledge" them. In this context, normative attitudes become relevant: our performances are not correct or incorrect according to various rules, but we can also "treat" them as correct or incorrect according to various rules. Autonomy is therefore related to the practice of assessing a performance as correct, but assessing is itself something that can be done correctly or incorrectly.

The dialogical competence that could characterize the autonomous agent could again be defined in semantic and normative sense. The main thesis we derive from the Brandom perspective is that the ones who engage in every discursive practice must distinguish between materially right or wrong inferences, where the adjective material indicates that the presence of some non-logical vocabulary is essential to the classification (i.e. we must know what is materially entailed in the use of our concept embedded in linguistic expressions). The participation in the game of giving and asking for reasons is possible by virtue of the fact that the participants master the normative vocabulary in terms of "commitments" and "entitlements" together with the relations among them (committive-inferences, permissive-inferences and material incompatibility).

The role of the autonomous agent indicates that when she is committed to some assertible content she takes responsibility to integrate it in a full set of commitments. At the same time, she draws the suitable consequences and put it under a rational criticism according to the criterion of material incompatibility. Again this point shows a misleading interpretation of the intersubjective linguistic practice. We must take into account the dimension of responsibility in a social context where the deontic attitudes of the agents play a fundamental role. On the one side, it is important to consider the contents of the reasons we advance in the game of giving and asking for reasons, on the other side, we must understand "how" an agent can take part in this game and give her contribution by undertaking and attributing responsibility and authority to her (and others') validity claims. Furthermore, if we relate only to the dimension of counterfactual robustness, it becomes difficult to explain the phenomenon of rational rectification when we meet reasons that do not belong to our community.
Even if we accept the inferential structure (based on material incompatibility) of this space proposed by Wilfrid Sellars, we must also give an explanation of the social perspectives from which we can undertake and attribute commitments. In this context, we have two possibilities: (a) to rely to the recognitional model presented by Brandom or (b) to rely to a kind of linguistic normative competence described in scorekeeping terms. In our opinion, this latter possibility is worthy of consideration as it addresses to a wide concept of justification of reasons for acting that does not require too strong conditions for autonomous agency.

According to (b), for an agent to be autonomous she ought to internalize the normative structure of a "dialogical" rationality. Because of the participation in the game of giving and asking for reason, we can master the communicative structure of justification by "default" and "challenge". Autonomy is relational in two senses: (1) the "semantic" ^[2] sense that shows the inferential commitments the agents must acknowledge and (2) the "pragmatic" sense that reveals the normative structure of that acknowledgment as a social net of deontic attitudes expressed by speech acts.

On our view, the autonomous agent plays the social role of scorekeeper, thus she is able or becomes able to justify her assertions. The entitlement to a claim can be justified by giving reasons for it, or by referring to the authority of another agent, or by demonstrating the capacity of the agent reliably to respond to environmental stimuli. The scorekeeping model is based on a notion of entitlement that presents a structure of "default" and "challenge". This model is fundamental in order to introduce autonomy as capacity of participation to the game of giving and asking for reasons. Which is the competence an agent must possess to be able to constitute an autonomous and critical voice in the public space?

Let us consider the case of a politician who is committed to the following action:«If the dissidents attack, I'll respond to». From the point of view of the justification by default, P could refer to his/her own knowledge of the norms that regulate war conflicts, or to the authority of others who are reliable. Naturally, this knowledge depends on the content of norms authorizing certain commitments. The fundamental trait of the scorekeeping model is that it represents a dynamic model, in which social practices are always exposed to the risk of dissent. In this context, social practices entail the dimension of "challenge", i.e. the case in which the scorekeeper challenges the interlocutor to justify and eventually to repudiate his/her commitment. The speech acts implied by this critical role are: disavowals, queries and challenges. Even in the case in which an agent acquires the entitlement to act by deferral, i.e. by indicating a testimonial path whereby entitlement to act can be inherited, query and challenge assume the function of fostering P's reflection. But if P can refer to the authority of a set of legal norms, it becomes difficult for the scorekeeper to alter the score of conversation. The disavowal is successful if the scorekeeper demonstrates to P that his/her inference implies incompatible entitlements: for example that the response to the attack entails catastrophic consequences, incompatible with moral commitments. In this case, P can be forced to perform a different inference such as: «If the dissidents attack, I'll find a diplomatic solution».