Introduction

James Stacey Taylor

In recent years, the concept of autonomy has become ubiquitous in moral philosophy. Discussions of the nature of autonomy, its value, and how one should respect it are now commonplace in philosophical debates, ranging from the metaphysics of moral responsibility to the varied concerns of applied philosophy. All of these debates are underpinned by an increasingly flourishing and sophisticated literature that addresses the fundamental question of the nature of personal autonomy.

   The concept of autonomy has, of course, been important for moral philosophy for some time, being central to the ethical theories of both Immanuel Kant and such contemporary Kantians as Thomas Hill and Christine Korsgaard.¹ However, recent interest in personal autonomy does not focus on the Kantian conception of autonomy on which a person is autonomous if her will is entirely devoid of all personal interests. Instead, it focuses on a more individualistic conception of this notion, whereby a person is autonomous with respect to her desires, actions, or character to the extent that they originate in some way from her motivational set, broadly construed.

   Interest in this individualistic conception of autonomy was stimulated by the publication of a series of papers in the early 1970s, in which Harry Frankfurt, Gerald Dworkin, and Wright Neely independently developed “hierarchical” accounts of personal autonomy.² The shared core of these accounts is both simple and elegant: A person is autonomous with respect to a first-order desire that moves her to act (e.g., she wants to smoke, and so she smokes) if she endorses her possession of that first-order desire (e.g., she wants to want to smoke). This approach to analyzing autonomy has much to recommend it. First, it captures an important truth about persons: They have the capacity to reflect on their desires and to endorse or repudiate them as they see fit. Second, it is an explicitly naturalistic and compatibilist approach to analyzing autonomy. As such, it fits well with the currently dominant compatibilist analyses of moral responsibility, and it seems able to disavow the implausible claim that personal autonomy is incompatible with the truth of metaphysical determinism – a disavowal that is defended by Bernard Berofsky and Alfred Mele in their chapters in this volume.³ Finally, this approach to analyzing autonomy is content neutral, for it does not require persons to hold any particular values in order for them to be autonomous. This enables it to be readily applicable to many debates within applied ethics where respect for autonomy is of primary concern and where this focus on autonomy is driven by the recognition that some means must be found to adjudicate between competing value claims in a pluralistic society.⁴

   Yet despite the many advantages of the hierarchical approach to analyzing autonomy, it suffers from significant theoretical difficulties. In the light of these criticisms, some proponents of the hierarchical approach to analyzing autonomy (such as Stefaan Cuypers and Harry Frankfurt) have developed sophisticated defenses of it.⁵ Other writers have developed a “second generation” of neohierarchical theories of autonomy that, while they move beyond the hierarchical approach to analyzing autonomy, acknowledge that the origins of their views lie in the original Frankfurt-Dworkin-Neely theory cluster. Two of the most prominent of these neohierarchical theories of autonomy are those developed by John Christman and Michael Bratman. Christman’s historical approach retains the hierarchical analyses’ requirement that the attitudes of the person whose effective first-order desire is in question are in some way autonomy conferring. However, rather than holding that this person must in some way endorse the desire in question for her to be autonomous with respect to it, Christman holds that she must not reject the process that led her to have this desire.⁶ Bratman’s analysis of autonomy – the key elements of which he outlines in the chapter “Planning Agency, Autonomous Agency” – combines his influential account of intention and planning agency with certain elements of the hierarchical approach to autonomy.⁷ Such neohierarchical approaches to personal autonomy have also been joined by a number of diverse and original approaches to analyzing autonomy that depart from the hierarchical approach altogether. These new approaches to analyzing autonomy include, but are not limited to, the coherentist approach of Laura Waddell Ekstrom,⁸ the “helmsman” approach of Thomas May,⁹ the doxastic approach of Robert Noggle,¹⁰ the sociorelational approach of Marina Oshana,¹¹ and the foundationalist approach of Keith Lehrer.¹² This debate over the nature of autonomy has led to a significant increase in the philosophical understanding of this concept, and so it is no longer correct that outside of the Kantian tradition autonomy “is a comparatively unanalyzed notion,” as John Christman was truthfully able to write in 1988.¹³ Moreover, the increasing attention that the concept of autonomy has recently received is not only of interest to autonomy theorists. This is because, as I outline in Section Ⅳ, which analysis of personal autonomy turns out to be the most defensible will have direct implications for all debates in moral philosophy in which this concept plays a major role.

   These, then, are exciting times for both autonomy theorists and all who draw upon the concept of autonomy. The chapters in this volume, each original to it, represent the state of the art of the current discussion of autonomy and the roles that it plays in discussions of moral responsibility and applied philosophy. The purpose of this Introduction, thus, is to provide the theoretical background against which these chapters were written, by outlining the progress of the debate over the nature and role of autonomy as this has unfolded over the past three decades. As such, it can naturally be divided into four sections. The first will provide the theoretical background to this collection as a whole, through outlining Frankfurt’s and Dworkin’s hierarchical analyses of autonomy together with the major criticisms that have led to their modification. Despite these modifications, however, I will note that even in their most recent forms these analyses are both still vulnerable to serious theoretical objections.

   The second section of this Introduction will outline three of the most prominent recent analyses of autonomy that have been developed to avoid the difficulties that beset the Frankfurt-Dworkin-Neely hierarchical approach: John Christman’s historical approach, Michael Bratman’s reasons-based view, and Laura Waddell Ekstrom’s coherentist analysis. The second section of the Introduction will serve as a supplement to the first, as it provides an introduction to the most recent theoretical literature on autonomy. In so doing, it will serve as a useful backdrop to the discussions in the first part of this collection, “Theoretical Approaches to Personal Autonomy,” in which Bratman and Ekstrom outline and develop their respective analyses of autonomy and in which the relationships among autonomy, free will, the “self,” and the concept of “identification” are considered.

   The third section of this Introduction will outline alleged connections between personal autonomy and moral responsibility. This will provide the theoretical background to the second part of this collection, “Autonomy, Freedom, and Moral Responsibility.” Finally, the last section of this Introduction will indicate the various ways in which the concept of autonomy is invoked within areas of contemporary philosophy apart from discussions of moral responsibility. This section will provide a useful basis from which to approach the final part of this book, “The Expanding Role of Personal Autonomy,” which focuses on the role that autonomy plays in political philosophy and in various fields of applied ethics.

1.  THE HIERARCHICAL ANALYSES OF AUTONOMY

The core feature shared by Frankfurt’s and Dworkin’s analyses of autonomy and identification is that these concepts are to be analyzed in terms of hierarchies of desire. (For the sake of clarity, I henceforth take the phrase “is autonomous with respect to her desire x” to be synonymous with the phrase “identifies with her desire that x.”)¹⁴ More specifically, on Frankfurt’s original analysis of autonomy a person is autonomous with respect to her first-order desire that moves her to act (her “will”) if she volitionally endorses that desire. (A “first-order” desire is a desire that a particular state of affairs obtains.) That is to say, a person is autonomous with respect to her effective first-order desire that x if she both desired to have the desire that x (i.e., she had a second-order desire that she have her desire that x, where a “second-order” desire is a desire about a first-order desire) and she also wanted her desire that x to move her to act (i.e., she endorsed her desire that x with a second-order volition).¹⁵ Similarly, on Dworkin’s original analysis of autonomy an “autonomous person is one who does his own thing,” where “the attitude that [the] person takes towards the influences motivating him…determines whether or not they are to be considered ‘his.’ ”¹⁶ That is to say, on Dworkin’s view a person is autonomous with respect to the desires that motivate him if he endorses his being so moved. In addition to requiring that a person’s motivations be “authentic” in this way, Dworkin also required that she enjoy both procedural independence and substantive independence with respect to her motivations. A person possesses procedural independence with respect to her motivations if her desire to be moved to act by them has not been produced by “manipulation, deception, the withholding of relevant information, and so on.”¹⁷ A person possesses substantive independence with respect to his motivations if he does not “renounce his independence of thought or action” prior to developing them.¹⁸

   On both Frankfurt’s and Dworkin’s hierarchical analyses, then, a person’s autonomy is impaired if she is moved to act by a desire that she does not volitionally endorse – if she has a second-order desire not to be moved by the first-order desire that is effective in moving her to act. In most cases, this is intuitively plausible. For example, if a person is subject to a constant neurotic compulsion to wash his hands from which he desires to be free, then his autonomy will be impaired if he is moved to act by a first-order desire to wash his hands that this neurosis causes him to have and by which he does not wish to be moved. Similarly, if a person is a “wanton,” if he does not care which of his desires moves him to act, then it seems plausible to claim that he is not autonomous (he is not “self-directed”), either because his “self ” is not engaged in directing his desires or actions or because he has no coherent “self ” to play this role.

   Yet despite their plausibility, these early hierarchical analyses of autonomy are subject to three serious objections. The first of these is the Problem of Manipulation.¹⁹ Frankfurt’s hierarchical analysis of autonomy is an ahistorical (or structural, punctuate, or time slice) account of autonomy, on which a person is autonomous with respect to his effective first-order desires irrespective of their historical origins, provided that he volitionally endorses them. The proponents of the Problem of Manipulation note that a third party (such as a nefarious neurosurgeon or a horrible hypnotist) could inculcate into a person both a certain first-order desire (e.g., the desire to smoke) and a second-order volition concerning this desire so that there is the pertinent sort of hierarchical endorsement. Because this inculcated first-order desire would satisfy Frankfurt’s conditions for its possessor to be autonomous with respect to it, Frankfurt is committed to holding that she is autonomous with respect to it – but this ascription of autonomy to her with respect to this desire is suspect.²⁰

   Of course, Dworkin’s analysis of autonomy is not directly subject to the Problem of Manipulation because it is blocked by his requirement that the process by which a person comes to have her desires be one that is procedurally independent – a condition that is clearly unsatisfied when a person’s desires are inculcated into her through hypnosis or neurosurgery without her consent. Despite this, one can still use the Problem of Manipulation to develop an indirect objection to Dworkin’s analysis of autonomy. Thus, although Dworkin’s requirement of procedural independence enables him to avoid the Problem of Manipulation, it only does so by fiat, by simply ruling ex cathedra that a person is not autonomous with respect to those desires that he has been manipulated into possessing. And this is not enough for his analysis of autonomy to be theoretically satisfactory. This is because an acceptable analysis of autonomy should not merely list the ways in which it is intuitively plausible that a person will suffer from a lack of autonomy with respect to her effective first-order desires, but must also provide an account of why a person’s autonomy would be thus undermined, so that influences on a person’s behavior that do not seem to undermine her autonomy (e.g., advice) can be differentiated from those that do (e.g., deception).

   Frankfurt’s and Dworkin’s analyses of autonomy also face the Regress-cum-Incompleteness Problem.²¹ On these analyses, a person is autonomous with respect to her effective first-order desires if she endorses them with a second-order desire. Because this is so, the question arises as to whether this person is autonomous with respect to this second-order desire and, if she is, why this is so. If she is autonomous with respect to this second-order desire because it is, in turn, endorsed by a yet higher-order desire, then a regress threatens, for the question will then arise as to whether she is autonomous with respect to this third-order desire – and so on. If, however, this person is autonomous with respect to the second-order desire for a reason other than its endorsement by a higher-order desire, then the hierarchical approach to analyzing autonomy is incomplete.

   Of course, the proponents of the hierarchical approach could avoid the Regress-cum-Incompleteness Problem simply by claiming that although the person in question is not autonomous with respect to her higher-order endorsing desire, she is autonomous with respect to her endorsed first-order desire, because autonomy is simply constituted by such an endorsement. Yet although Frankfurt and Dworkin could avoid the Regress-cum-Incompleteness Problem by adopting this line of response, neither of them does so, no doubt because they recognize that were they to do so they would encounter the equally troubling Ab Initio Problem: How can a person become autonomous with respect to a desire through a process with respect to which she was not autonomous? Or, in other words, how is it that a person’s higher-order desires possess any authority over her lower-order desires?²² When put in this way, the Ab Initio Problem is often termed the Problem of Authority and in this guise has been neatly encapsulated by Gary Watson: “Since second-order volitions are themselves simply desires, to add them to the context of conflict is just to increase the number of contenders; it is not to give a special place to any of those in contention.”²³

   Faced with these three difficulties, both Frankfurt and Dworkin modified their original analyses. Recognizing that his analysis would be subjected to the Regress-cum-Incompleteness Problem, Frankfurt attempted to eliminate the possibility of such a problematic regress by claiming that a person’s decisive identification with one of his desires would terminate it.²⁴ Frankfurt elaborated this decision-based version of his hierarchical analysis of autonomy in “Identification and Wholeheartedness,” where he argued that a person is autonomous with respect to his effective first-order desire if he decisively endorses it with a second-order volition. Directly responding to the Regress Problem, Frankfurt claimed that if a person endorses his effective first-order desire “without reservation . . . in the belief that no further accurate inquiry would require him to change his mind,” it would be pointless for him to continue to assess whether he was autonomous with respect to the first-order desire that was in question.²⁵ Furthermore, a person’s decisive identification with his endorsing second-order volition also seems to circumvent the Ab Initio Problem/Problem of Authority, for through this decision the person in question will endow his volition with the authority that it previously lacked.

   Unlike Frankfurt, Dworkin did not directly attempt to address criticisms of his analysis of what conditions must be met for a person to be autonomous with respect to her desires and actions. Instead, he clarified that his account was concerned not with the local conception of what conditions must be met for a person to be autonomous with respect to her actions (or desires), but, instead, with a more global conception of autonomy as a “second-order capacity of persons to reflect critically upon their first-order preferences, desires, wishes and so forth.”²⁶ Dworkin argued that once it is understood that he was not trying to provide an account of what made a person autonomous with respect to her desires or actions, his conception of autonomy avoids the Regress-cum-Incompleteness Problem. This is because, he claimed, as long as a person enjoyed procedural independence with respect to her reflection upon her desires, there would be “no conceptual necessity for raising the question of whether the values, preferences at the second order would themselves be valued or preferred at a higher level.… ”²⁷ Similarly, Dworkin held that his account of autonomy is unaffected by the Ab Initio Problem/Problem of Authority. Because on his view persons enjoy autonomy when they engage this capacity for reflection, the exercise of this second-order capacity for endorsement just is what is involved in being autonomous.

   Yet even if Dworkin’s more global approach to analyzing personal autonomy avoids the major problems that were outlined above, this is achieved at considerable cost. This is because in many discussions that concern the nature of autonomy the issue is not what psychological capacities a person must possess to have the capacity for autonomy, for it is generally accepted that to be autonomous an agent must possess the ability to engage in some form of second-order reflection of the sort that Dworkin outlines. Instead, what is really of interest in discussions of autonomy is the question of how the exercise of this psychological capacity for reflection results in persons being autonomous with respect to their desires and actions. Thus, in adopting this more global approach to autonomy Dworkin is no longer offering an analysis of autonomy that is congruent with the discussions in moral philosophy in which autonomy plays a major role, for these discussions focus on the more localized question of what makes a person autonomous with respect to her particular desires or her particular actions.

   Once Dworkin’s more recent aims in developing an analysis of autonomy have been clarified, then, they can be seen to be distinct from the primary aim of most autonomy theorists – namely, to provide an account of what it is for a person to be autonomous with respect to her desires and her actions. Yet this core aim of autonomy theorists is not satisfied by Frankfurt’s decision-based analysis of autonomy either, for it fails as a successful response to three of the objections outlined above. First, the mere fact that a person has decisively identified herself with a particular first-order desire does not halt any possible problematic regress. This is because, as Frankfurt later recognized, the Regress-cum-Incompleteness Problem would still arise, given that one could still question whether the person in question was autonomous with respect to this decision. Furthermore, the Problem of Manipulation still poses difficulties for this account because such a decision could still be the result of the agent’s succumbing to forces that are external to her. For example, she might have been hypnotized into decisively identifying with a given desire.²⁸ Finally, because a person can be manipulated into decisively identifying herself with a particular first-order desire, the proponents of the Ab Initio Problem/Problem of Authority can still question why such mental acts are authoritative for her.

   Frankfurt recognized that his analysis of autonomy was beset by these three problems because it rested on the claim that a person became autonomous with respect to her desires through endorsing them with a “deliberate psychic event” – and one could always question whether the person in question was autonomous with respect to this event. To avoid these criticisms, Frankfurt developed a satisfaction-based analysis of identification.²⁹ On this analysis, a person need not engage in any “deliberate psychic event” for her to identify with her desires. Instead, on this analysis a person is autonomous with respect to a desire if he accepts it as his own – if he accepts it as indicating “something about himself.”³⁰ In accepting a desire, a person will reflect on it to see if it is expressive of something about him. If it is, then he will form a higher-order attitude of acceptance toward it as part of himself. It is this acceptance of the desire that constitutes the person’s endorsement of it, to use Frankfurt’s “misleading” terminology from “The Faintest Passion.”³¹ The sense of endorsement that Frankfurt is using here, then, is the sense in which one might endorse the claim of an entity to be a member of a class, without thereby evaluating (either positively or negatively) the merits of the particular entity that is making the claim. Once a person has met the requirement that she reflectively endorse her first-order desires in this way, Frankfurt does not also require that she then reflectively endorse her attitude of endorsement, for, as he rightly notes, such a requirement would lead to a regress. Instead, Frankfurt holds that a person will identify with a first-order desire if she is satisfied with the higher-order attitude of endorsement (i.e., acceptance) that she has taken toward it. For Frankfurt, a person’s being satisfied with his attitudinal set “does not require that [he] have any particular belief about it, or any particular feeling or attitude or intention.… There is nothing that he needs to think, or adopt, or to accept; it is not necessary for him to do anything at all.” Instead, his being satisfied with his attitudinal set simply consists in his “having no interest in making changes” in it.³² And this, notes Frankfurt, is important, for it explains why this analysis of identification as satisfaction is not subject to a problematic regress of the sort that beset his earlier analyses.³³ Here, then, a person will be autonomous with respect to his effective first-order desire if he is not moved to make changes in his motivational economy when he is moved to act by it, if he is satisfied with it.

   Frankfurt’s satisfaction-based analysis of autonomy is not subject to the Regress-cum-Incompleteness Problem for the reasons outlined above. Moreover, it is also not subject to the Ab Initio Problem/Problem of Authority. This is because Frankfurt has now clarified that a person’s higher-order attitude of acceptance toward her lower-order desires does not possess any normative authority over them; instead, these attitudes are merely used by the person in question to assess whether her lower-order desires are to be regarded as being descriptively hers, whether they flow from her (broadly Lockean) self. However, this analysis of autonomy still faces the Problem of Manipulation. This is because a person could unwittingly be hypnotized into possessing a certain first-order desire in such a way that he believes that it originates from within him. Given this belief, he would then both endorse this first-order desire and be satisfied with it, in Frankfurt’s senses of these terms. This person would thus meet all of Frankfurt’s most recent criteria for him to identify with his hypnotically induced desire – yet surely such a desire is one with respect to which its possessor is paradigmatically heteronomous.

II.  NEW APPROACHES TO AUTONOMY

Christman’s Historical Analysis

From the previous discussion, it might seem that the hierarchical approach to analyzing personal autonomy is doomed to failure, in large part because it appears inevitably to succumb to the Problem of Manipulation. Yet this assessment of hierarchical theories of autonomy needs to be qualified, for the focus of the past discussion was on Frankfurt’s explicitly ahistorical approach to analyzing autonomy. Recognizing the difficulties that such an approach would have when faced by the Problem of Manipulation, Christman developed an explicitly historically based version of the hierarchical approach to analyzing autonomy. For Christman, an agent P is autonomous relative to some desire (value, etc.) at time t if and only if

1. P did not resist the development of D (prior to t) when attending to this process of development, or P would not have resisted that development had P attended to the process;
2. The lack of resistance to the development of D (prior to t) did not take place (or would not have) under the influence of factors that inhibit self-reflection;
3. The self-reflection involved in condition i is (minimally) rational and involves no self-deception;³⁴

and

   Unfortunately, as it stands, Christman’s historical analysis of autonomy fails to provide either necessary or sufficient conditions for a person to be autonomous with respect to her desires. To see that this account does not provide necessary conditions for a person to be autonomous with respect to her desires, imagine a child at time t whose mother wished him to learn to play the piano and who beat him if he did not practice.³⁶ As time passes and the child grows more proficient at playing, he discovers (at time t1) that his mother’s belief that piano playing suited him was right, and he comes to love playing – even though he still repudiates the means by which his mother brought him to this position. Thus, even though at t1 this person rejects the process by which he was brought to desire to play the piano, at t1 (and onward) he appears to be fully autonomous with respect to this desire.³⁷