Introduction
Michael Bratman’s hierarchical model of self-governance is one of the most influential and thorough accounts of autonomous agency in the current debate on the philosophy of action.1 It aims to be an improvement on previous hierarchical accounts (especially Harry Frank- furt’s) in that it tries to answer common objections to them. But it also goes beyond those accounts insofar as it adds features not available to them. More specifically, it is by incorporating into the core of his account of autonomy a temporally extended sense of agency (which involves “psychological conti- nuities” constrained by rational demands for consistency, coherence and sta- bility) that Bratman makes his account remarkable.
Our task here is twofold: to present a detailed account of Bratman’s self-governance model of autonomy and assess its plausibility. In assessing its plausibility, we will raise objections to it and try to show that Bratman’s account faces serious difficulties: on the one hand, in making autonomy entirely depend on higher-order self-governing policies and, on the other hand, in understanding temporally extended agency essentially in terms of rational demands for consistency, coherence and stability.
Before we go through the details of Bratman’s account, we should first introduce the problem of autonomy as understood by the tradition of hierarchical accounts from which Bratman departs.
What autonomy is about
is usually said that a complete theory of freedom must provide answers to two questions. One question is metaphysical, the other is psychological and practical. The first question is about the principle of alter- native possibilities (its meaning and cogency). The second question is about autonomy, which involves centrally a question about explaining agential au- thority in practical reasoning, deliberation, motivation and action. It is deba- table how those two questions relate to each other. Nevertheless, it seems conceptually possible to discuss them (at least in part) independently of one another, or so we will be assuming here.2 In what follows, we will be speci- fically concerned with the psychological and practical problem of autonomy.
The problem of autonomy is traditionally characterized as the problem of the determination of the will. There are, of course, many theories of the will, whose answers to that problem shall vary greatly. It is not part of our aim here to explore this variety. Here we are particularly interested in how a tradition of hierarchical accounts (especially those inspired by Harry Frankfurt’s writings) tackle the problem of autonomy. For this tradition of hierarchical accounts, the problem of the determination of the will is one about the possibility of the will having a certain psychological functioning and structure that is in part moulded by the agent.3 Accordingly, whenever the agent is able to determine her own will, the agent determines the psy- chological functioning of her own deliberations, motivations and corres- ponding actions through practical reflection and reasoning by their being embedded in a certain psychological structure. In this sense, those psycholo- gical elements can be said to exhibit agential authority.4 This means that the agent is, as it were, “fully behind them”: their functioning and the psycholo- gical structure in which they are embedded are the expression of the agent’s practical standpoint and, thus, of herself as an agent.
According to this tradition of hierarchical accounts, the basic idea of autonomy is best captured by the metaphor of stepping back from a mo- tivational psychological element (say, a pro-attitude or desire, understood broadly),5 reflecting on and asking whether to act on it. There are two moves in trying to make sense of this question (and metaphor): one is about the structure of our practical thought, the other is about its authority.
From a structural perspective, the question about whether to act or not on a desire is naturally understood as inviting a response from a hi- gher-order perspective, which makes conceptual room for a hierarchical account. One needs to know from such a higher-order perspective how to manage hierarchically the desires that make motivational pressure on one to act. So, one part of the answer of hierarchical accounts to the problem of autonomy is given in terms of the reflective structure of our practical point of view. Hierarchical accounts claim that the central way in which we reflec- tively manage our practical life is inherently hierarchical.
Besides, we also need an explanation of how to answer the reflective question about the satisfaction of a desire. On hierarchical accounts, this is often put in terms of a question about identification with a desire. In asking for identification with a desire (and the ensuing course of action), the agent asks from a higher-order perspective whether that desire is something to be endorsed and acted on so that it is fully entitled to “speak for the agent”. This is a question about the authority a desire (and its corresponding action) bears from the agent’s practical standpoint. Here hierarchical accounts vary in their answers. A hierarchical account can hold that identification is cons- tituted by a second-order or higher-order desire, decision, intention, policy, reason-giving judgment, value judgment, etc.6 Hence, for hierarchical ac-
counts of agential authority (and autonomy), the question of Identification is about the specific functioning of the (type of) psychological elements which are constitutive of the agent’s hierarchical structure of reflective practical thought.
Now, two comments are in order. Firstly, hierarchical accounts aim to provide a “non-homuncular” explanation of autonomy (Bratman 2007g: 177; Bratman 2007i: 201 ff.). This means that the autonomous agent is not to be understood as existing independently of her psycho- logical materials, functioning and structure. Actually, for hierarchical ac counts, autonomy is a property to be specified in terms of those psycho logical elements. The autonomous agent is to be defined in terms of them. Secondly, agential authority is a property of psychological states and their functioning (which are embedded in a certain structure). And the key notion to explain agential authority is identification with a desire (and its inherent relation to action).
Frankfurt’s versión of the hierarchical account
I n order to better understand Bratman’s self-governance account of autonomy, it will be instructive to go through some of the details of Harry Frankfurt’s views, given that this is a departure point for Bratman.
On Frankfurt’s initial view, one’s being capable of autonomy is just a matter of one’s having a second-order desire (or, as Frankfurt prefers to call it, a second-order volition) about one’s first-order desire (Frankfurt 1988b: 16). Frankfurt points out that we are able not only to have a desire towards some course of action but also that we are able to reflect on our first-order desires. In so doing, we are supposed to be able to reflect on which desire we want to constitute our will, i.e. which desire we want to act on. Since Frank- furt’s initial view seems to suppose that motivation to action is necessarily tied ultimately to desires, any reflective question about whether to act on first-order desires will inevitably receive an answer from another desiderative attitude, now from a second-order perspective. Such a capacity for reflection and second-order desiring enables Frankfurt to state the conditions under which an agent is autonomous: an agent is autonomous when she acts out of a desire she desires to desire. In this sense, she can be said to identify (from a second-order perspective) with a first-order desire and such desire (and ensuing action) bears agential authority. Let us follow Bratman in calling this kind of ability weak reflectiveness (Bratman 2007b: 23).
So, according to Frankfurt’s account, an autonomous agent ex- hibits at least two features that a non-autonomous agent lacks (or does not exercise or fails in exercising), namely, reflection and second-order desi- ring.7 In this sense, even if we were to grant that the actions of the agents who lack (or do not exercise or fail in exercising) those two features are somehow purposive or intentional (because they are appropriately related to the agents’ first-order desires) they would not count as autonomous. The agents could or would not take a stand on their motivations and reflectively govern their actions.
Having said that, we should now ask: is weak reflectiveness appropriate (necessary and sufficient) to explain autonomy? It seems not. Let us see why.
Gary Watson (1975: 217 ff.) raises a strong objection to Frankfurt’s account. Watson simply calls into question the plausibility of a hierarchical account of desires as providing the adequate understanding of agential au- thority. The problem, Watson argues, is that there is no reason why iden- tification with a desire should occur on a second-level basis. After all, on Frankfurt’s account, second-order agency is constituted, just like first-order agency, simply by the functioning of desires. Watson then claims that, if se cond-order desires are simply desires, why should they have “authority” over first-order desires? As Watson puts it, it seems that Frankfurt’s strategy ends up increasing the number of contenders without justification.8 The problem might go deeper if we ask how any higher-level attitude could do the job, in which case Frankfurt’s proposal should face a threat of regress.
Frankfurt himself seems to have agreed with Watson’s challenge that an appropriate characterization of autonomy (agential authority and identi- fication) should go beyond weak reflectiveness given that he has since then made many modifications to his initial view.9 Here we will focus on two further conditions that Frankfurt has added (at different moments in his works) to his initial view in order to meet Watson’s challenge.
Frankfurt proposes to add to his initial view:
a higher-order decisión;
a state of satisfaction (with a given psychological configuration or state of affairs).10
Let us now see these two conditions in more detail. They will be im- portant for our following discussion, since Bratman partly develops his own account as an improvement and qualification on those conditions.
Frankfurt’s qualified versión of the hierarchical account
Frankfurt’s attempt to understand autonomy as involving a kind of decision is amendment to his initial view. On this amended version, identification with a desire involves a decision an agent makes on a hi gher-order basis concerning his second-order desires (hence also concerning her first-order desires). The agent decides to make a given second-order desire her will. So, the exercise of autonomy still has a higher-order structure, but it draws on a different psychological element, namely, a decision. At first sight, if identification involves a kind of decision, it might be plausible to maintain that such proposal would avoid the objection of threat of regress, since it makes no sense to doubt whether an agent’s decision is her own. An agent’s decision is supposed to be something she really makes (in opposition to it simply “occurring” in her), something she cannot be alienated from or disown. As Frankfurt puts it “decisions, unlike desires or attitudes, do not seem to be susceptible both to internality and to externality”-that is, to a question whether they are constitutive of the agent’s own practical stan- dpoint or not (Frankfurt 1988c: 68).
Notwithstanding Frankfurt’s attempt to establish a necessary con- nection between the ownership of a decision and agential authority (and then identification), David Velleman raises a forceful objection against it. While considering the possibility of some sorts of decisions that may not be entirely transparent to one’s mind when they are made and that can only be thoroughly accessed retrospectivelyVelleman claims that:
I may conclude that desires of mine caused a decision, which in turn caused the corresponding behaviour; and I may acknowledge that these mental states were thereby exerting their normal motivational force, una- betted by any strange perturbation or compulsion. But do I necessarily think that I made the decision or that I executed it? Surely, I can believe that the decision, though genuinely motivated by my desires, was thereby induced in me but not formed by me; and I can believe that it was ge- nuinely executed in my behaviour but executed, again, without my help. (Velleman 2000: 126-127)
Velleman’s counterexample rests on the idea that there are cases of motivated decision that lack the distinctive mark of agential authority and then identification. In this respect,Velleman is suggesting that a decision is a psychological element which is prone to the same difficulties as desires with respect to autonomy. Accordingly, we can (if not always, at least on a suffi- cient number of occasions) ask for the authority of a decision.10
In face ofVelleman’s challenge, Frankfurt could simply say that agential authority (and identification) is restricted to conscious and throughout trans- parent to one’s mind decisions. But this response would sound ad hoc. Besides, we can imagine cases where making a conscious and transparent to one’s mind decision is not sufficient for identification. For example, an unwilling addict could decide to take a drug because he realises that he cannot resist the desire to take it and does not want to suffer from the side-effects of abs- tinence in his mind and body. Drawing on a useful expression by Cullity and Gerrans (2004: 325), the unwilling addict might arrive at a “merely desire-re- moving” decision and become a self-managing addict. Still, it seems plausible to say that, although he has become a self-managing addict and his decision to take the drug is conscious and transparent to him, the decision is not one with which he is identified. He is, in the face of contextual pragmatic pres- sures, somehow “forced” to make such a decision.
In an attempt to overcome that result, Frankfurt introduces the idea of“satisfaction”:
Being genuinely satisfied [...] is a matter of simply having no interest in making changes. What it requires is that psychic elements of certain kinds do not occur. But while the absence of such elements does not require either deliberate action or deliberate restraint, their absence must nonetheless be reflective. (Frankfurt 1999a: 104-105)
Accordingly, identification with a desire is just a matter of being sa tisfied (from a reflective higher-order perspective) with a kind of psycho- logical configuration or state of affairs, where “the agent is satisfied” means something like “to be reflectively aware of her second-order desires con- cerning her first-order desires and not intend to make any changes or leave thing as they are”. This is, as Bratman puts it, not an attitudinal feature, but a sort of “structural feature of the psychic system” (Bratman 2007i: 203; Bratman 2007e: 94; Frankfurt 1999a: 104).
However, on the face of it, the proposal is not free of problems as it all depends on how to understand “non-occurrence of psychic ele ments”. Frankfurt wants to make sure that “non-occurrence” does not mean “choosing to leave things as they are” (Frankfurt 1999a: 104). That is to say, Frankfurt wants to reject the idea that such a state of satisfaction involves a further choice or decisión to leave things as they are. Instead, he wants sa tisfaction to be more like “a state constituted just by the absence of any tendency or inclination to alter its condition” (Frankfurt 1999a: 104). But Frankfurt also needs to distinguish this from a simply unreflective state of lack of concern. So, the question becomes whether it is possible to interpret satisfaction as pointing somewhere between a decision and lack of reflective concern. Alas, this proves to be a hard question and an answer to it does not come straightforward.
One way of interpreting satisfaction along those lines would be to say that an agent’s wanting “to leave things as they are” involves the presence in her mind of something similar to either a feeling of well-being or a fe- eling of quietude with a state of affairs.11 Either way, the interpretation is problematic, as it does not seem to preclude cases in which an agent thinks she identifies with a desire by feeling contentment or quietude, but which may have on many occasions, so to speak, a dubious origin. After all, those feelings can be induced by states like euphoria, anxiety, depression, ener- vation, and so on, which are most naturally interpreted as apt candidates for hindering an agent from an accurate and reliable assessment of her own situation. The point here seems to be that those feelings are by their very nature too unstable, prone to all sorts of influences, and subject to a huge va- riety of causes and effects, so that it is very unlikely that they would reliably play the role that identification with a desire is supposed to play in making it central to autonomy.
Another natural way of understanding such a state of satisfaction would be to understand “non-occurrence of psychic elements” as meaning “decide not to decide to change things as they are”. At first sight, this would not in- volve a direct decision to leave things as they are, as Frankfurt warns us against taking this path. But, as should be clear, the solution is of no help for Frankfurt, given that it would anyway reintroduce another kind of decision into the picture (and make it again subject to the objections we have seen before).12
Bratman’s hierarchical account of autonomy as self- governance
B ratman offers a model inspired by Frankfurt’s account that pur- ports to be immune to the criticisms raised above. Bratman agrees with many objections raised against Frankfurt’s account, but also thinks that there is something right in Frankfurt’s view. More specifically, he favours a hierarchical structure and, similarly to Frankfurt’s amended version, he claims that identification involves typically a kind of decision. Finally, Bratman also thinks that a state satisfaction is a necessary element of autonomy.
However, for Bratman, simply taking a decision is not sufficient for autonomy, nor is satisfaction with a decision a mere experience of con- tentment or quietude or a state restricted to a certain psychological con- figuration. According to Bratman (2018; 2007), autonomous agency is part of a larger planning or policy-oriented structure that involves a temporally extended exercise of agency13 and is subject to rational demands of consistency,
coherence and stability.14 By their being embedded in such rationally cons- trained temporally extended sense of agency, autonomous decisions are typically general and normative.15 As such, they get expressed in the form of intentions, plans and policies for taking certain desires as justifying conside- rations for practical reasoning, deliberation and action. Satisfaction is to be understood on these grounds as well. For Bratman, satisfaction is a matter of the stability of consistent and coherent relations among intentions, plans and policies. Let us see this account in more detail.
Humans are planning creatures (Bratman 1987; 1999; 2007; 2018).
They can settle on complex, future-directed and partial plans that shape future practical reasoning, deliberation and actions by playing a role in cross-temporal coordination and organization of the agent’s practical life.
These plans are partial because they are very general; they are complex be- cause they involve and affect a whole set of motivational attitudes, practical judgements, deliberative processes, beliefs and actions through the time; they are future-directed because they do not restrict themselves to action that is performed just from moment to moment. So, settling on a plan involves a kind of commitment. In virtue of constraints of rational and pragmatic pressure, plans involve stability. Although a plan can always be abandoned in the light of new information, there is a rational and pragmatic pressure for not reconsidering or abandoning some prior plan. The content of plans | 159 point to some general aim or end to be pursued and they further impose ra- tional demands on future practical reasoning and action such as consistency and means-end coherence. Intentions, plans and, more broadly, policies are psychological elements that exhibit a planning structure.
At first sight, it is not clear how a planning agent in Bratman’s terms may be conceived without appealing in advance to an actual property of the
agent’s being extended over time. It seems that in order for it to be possible for an agent to be apt to settle on a plan, the agent must have previously some basic notion of herself as integrated through time and as having al- ready developed cross-temporally connected capacities for reflection, rea- soning and action. Thus, it is not entirely clear that the planning structure of agency may be conceived as logically independent of the agent’s own previous understanding of herself as being a temporally extended agent and of her actually being temporally extended. But then it is not clear whether it is intentions, plans and policies that give rise to rational demands of consis- tency, coherence and stability over time or whether it is a more fundamental (metaphysical) understanding of agency (and the agent’s identity) as being temporally extended and subject to rational demands that explain why in- tentions, plans and policies might fit in. However serious it may be, we are not going to pursue this difficulty here.16
Be that as it may, what is important about temporally extended agency, according to Bratman, is that the agent is capable of seeing her ac- tivities over time as performed by her, as being initiated and finished by herself. This capacity of connecting a whole web of psychological elements and actions through time is what makes possible the functioning of planning structures as coordinating and organizing one’s practical life. It is then the exercise of those capacities structurally organized in a cross-temporally planning framework of agency that supports the psychological connections and continuities that Bratman calls “Lockean ties” (Bratman 2007b: 29 ff.). Just as Lockean accounts of personal identity, Bratman also thinks that psy chological connections and continuities over time are “constitutive of the identity of the agent over time, an identity that is presupposed in much of our practical thinking” (Bratman 2007a: 5). Thus, it is primarily its role in constituting and supporting this organized, cross-tem- poral, Lockean interweave of action and practical thinking that confers on a structure of attitudes a claim to speak for the agent-a claim to agential authority. (Bratman 2007a: 5)
Bratman then finds in the notion of higher-order policies the best can- didate for playing that coordinating role of agency over time. Higher-order policies, unlike singular intentions and plans-which are typically related to particular moments or are temporally restricted-are more general and are concerned with how to treat motivational states like desires over time from the agent’s own perspective as temporally extended. Thus, by having desires as their primary objects, they are structurally hierarchical. More specifically, they endorse or support a desire by either treating it as a justifying or a non-justifying consideration (or reason-giving or end-setting) in one’s motiva- tionally effective practical reasoning and deliberation (and ensuing action).17 So, such policies basically come in two forms: they can be pro or con a desire and its associated practical reasoning, deliberation and action. Bratman calls these higher-order policies “self-governing policies”.
So understood, self-governing policies are what explain identification with a desire. To identify with a desire is to decide to treat the desire as rea- son-giving, which then involves settling on a higher-order general self-go- verning policy with that content. More precisely, for Bratman, to treat a desire as reason-giving is to treat it as a consideration under which the agent can justify the performance of relevant means to achieve it as an end over time, and which then is embedded in a cross-temporal structure that satisfies the rational demands of consistency, coherence and stability. Self-go- verning policies are then psychological elements that by their very nature, structure and rational connections constitute identification with desires over time and confer agential authority on them, as well as on associated practical reasoning and action. In this sense, self-governing policies shape an agent’s practical life by way of supporting psychological connections (or Lockean ties) that constitute and sustain an agent’s own personal and practical identity as temporally extended.
Examples of such policies are: “doing exercises regularly”, “being to- lerant of others”, “never smoking in public places”, “never treating a desire for revenge as a consideration in my deliberative practical reasoning”, etc. Although the formulation of the policy may not always make explicit re- ference to desires, Bratman is assuming (as a corollary of the hierarchical account) that all policies can be rephrased in such a way that they involve a reference to a desire as their primary object. Thus, a policy like “never smoking in public places” could be rephrased as a policy for not treating the desire to smoke in public places as reason-giving (or for treating it as a non-reason-giving). The policy of doing exercises regularly might be rephrased as the policy for treating the desire for exercising regularly as reason-giving, and so on.18
Hence, self-governing policies are essentially general normative com- mitments; commitments to self-management as a solution to the practical problem that the motivations typical of human agency give rise to by way of reflection. Self-governing policies are a species of normative guidance instructions about how to manage an agent’s practical life cross-temporally. In this sense, self-governing policies not only sustain an agent’s practical identity over time, they also play a role in constituting and creating it.
Those are the basic tenets of Bratman’s self-governance account of autonomy. Before we complete its presentation, some further clarifications are needed in order to avoid some natural objections to it.
At first sight, there is something missing in the picture. It seems that the self-governing policies themselves need some kind of endorsement. In having a self-governing policy that supports treating a desire as rea- son-giving the agent decides to do so. But then it seems always possible to ask for the authority of such policies and their corresponding decisions. As we have seen, it seems pointless to insist on another level of thinking, as this would amount to reviving the threat of regress. Bratman’s suggestion at this point is to draw on the notion of satisfaction by giving to it contours different from Frankfurt (Bratman 2007a: 6, n. 6). Unlike Frankfurt’s sug gestion-i.e. of understanding satisfaction as an experience of contentment or quietude or as an unspecified favourable state directed at a particular psychological configuration-, Bratman’s understanding of satisfaction is directly tied to the rational organizing and coordinating role of self-go verning policies in terms of consistency, coherence and stability through time. For Bratman, satisfaction with a self-governing policy means “not to have another policy that challenges it”, that is, satisfaction is understood in terms of absence of conflict among policies (Bratman 2007b: 34 ff.). Thus, in this sense, contrary to Frankfurt, satisfaction is a structural feature of agency (and not merely of the psychic system) and understood in terms of the cross-temporal structure of agency. For Bratman, satisfaction is not particular or episodic, nor is it directed at a psychological configuration. Satisfaction is the result of the role an agent’s self-governing policies play cross-temporally in the agent’s practical life. So, satisfaction is a kind of su- pervenient property on the realization and execution of an agent’s self-go- verning policies over time.19
Now we can see how Bratman’s understanding of satisfaction as a function of the role that self-governing policies play in an agent’s life over time gives him resources to answer objections such as Watson’s and Velle- man’s. Bratman can give a unified response to both. What confers agential authority to a psychological element and stops the potential regress to which Watson draws attention is exactly the fact that self-governing policies typi- cally involve decisions. But the decisions involved in self-governing policies are immune to the kind of objection raised by Velleman because they are not episodic; they involve the agent’s normative commitments as a temporally extended being. So, decisions like those ofVelleman’s example (not entirely conscious or transparent to one’s mind) are not representative of an agent’s practical and personal identity in that they are not the expression of self-go verning policies and their normative commitments, which are ultimately an expression of an agent’s personal and practical identity.20
Another needed qualification concerns the fact that it seems possible, on some occasions, that the agent’s reflectively treating a desire as reason- giving might occur quite independently of the self-governing policies the agent holds (Bratman 2007g: 179-185). That is to say, on the one hand, treating a desire as reason-giving is something that might reflectively occur in the absence of a policy supporting it. On the other hand, on some occasions, the linkage between the execution of the policy and the functioning of the particular practical reasoning might turn out to be severed.
Arguably, in cases like these, the functioning of the practical reasoning process is non-autonomous. Although the agent goes through reflection and practical reasoning, this functioning may not bear agential authority because it does not occur by way of the execution of a policy and, in this sense, it is not fully governed by the agent. Drawing on an expression coined by Gi- bbard, Bratman says about those cases that, although motivationally effective practical reasoning occurs, “the agent is not governing the reasoning, but is instead in the ‘grip’ of concerns that drive the reasoning” (Bratman 2007g: 182). This sort of case forces Bratman to make the linkage between self-go- verning policies and actual practical reasoning processes tighter.
Bratman copes with that by introducing two further conditions, namely, reflexivity and transparency (2007g: 184 ff.; 179 ff.). Reflexivity re- quires that treating a desire as reason-giving occurs by way of the very policy that supports it. The transparency condition, in turn, requires that the agent knows of this linkage. Thus, reflexivity introduces a self-referential condition on the execution of a self-governing policy; transparency introduces a self- knowledge condition.
We have now completed the presentation of Bratman’s machinery, his account of autonomy as self-governance and the kind of “strong form” of reflectiveness and agency that it involves. We have seen that autonomous action, for Bratman, is action performed by (or under the guidance of) self-governing policies which constitute and make intelligible (through ra- tional demands of consistency, coherence and stability) an agent’s treatment of her desires as justifying considerations in practical reasoning and delibe- ration over time. Desires that exhibit those features are the ones with which the agent identifies and, as such, they bear agential authority.
Troubles for Bratman proposal that autonomous agency is better characterized in terms of properties (including structural properties) of the psychological states of the agent (thus avoiding a “homuncular” account) is undoubtedly worthwhile. Also, the idea that the exercise of autonomy is not episodic or, as it were, atomistic seems to be pointing in the right direction. Similarly, the thesis that autonomy cannot be understood in- dependently of what is constitutive of the agent’s practical and personal identity through time deserves attention. Finally, making rational pressures for consistency, coherence and stability as a requirement for the realization of autonomy is also something, as far as it goes, quite compelling. After all, inconsistent, incoherent and fickle agents surely exhibit some kind of
failure (although it may not be entirely clear in which sense it is a failure of rationality).21
Still, there seems to be something odd with Bratman’s account. On the one hand, it is too restrictive. It does not seem to be true that all there is (or even centrally) to autonomy is a matter of having higher-order policies held and executed in a rationally constrained temporally extended structure of agency. There seem to be many aspects of our practical lives that have a bearing on autonomy but do not require a policy-oriented structure. On the other hand, Bratman’s account seems to make autonomy come out too cheap. Accordingly, by simply being able to organize our practical lives (in terms of deliberations, motivations and actions) consistently, coherently and stably over time, we get everything that is worthy of the name of autonomy. Nevertheless, organizing our practical lives rationally (consistently, cohe rently and stably), although undoubtedly a necessary element in making our lives possible, may be far from the whole story about autonomous agency. Dealing rationally with practical matters may have a purely pragmatic di- mension and, as such, leave out many aspects that may be relevant for an adequate view on autonomy.
This sort of critical diagnosis indicates that, although Bratman’s ac count may have a number of virtues and deep insights, it fails as a complete or overall compelling account of autonomy. It fails, in particular, at providing necessary and sufficient conditions for autonomy.22 We will now explore this diagnosis by offering two general criticisms. Both of them point in the di- rection that higher-order policies understood as essentially involving rationally constrained temporally extended agency fall short of capturing central intuitions about autonomy and give rise to an unstable account.
6.1. Self-governing policies are insufficiently specific
We have seen that Bratman’s account provides two modalities in which self-governing policies relate to a desire, namely, as a pro-policy or con-policy. This is non-problematic. But the conclusions Bratman draws from it-particularly, how the policy relates to the contení of desires-are misleading. Bratman holds that identifying or non-identifying with a specific type of desire is having, respectively, a pro-policy or a con- policy towards it (Bratman 2007b: 28; 33 ff.; 2007c: 60 ff.; 2007d: 69 ff.). So, self-governing policies have as their contents specific types of desires (whose tokens may recur over time). Whenever the agent has a particular token desire of the type of desire specified by the content of the self-go verning policy, the policy works as a normative guidance about what to do with it (either to treat it as reason-giving or not). However, this is a too simplistic description of how self-governing policies should work, one that seems insufficient and incompatible with the role Bratman ascribes to them in organizing and coordinating the whole of our practical lives. If Bratman’s description of how self-governing policies work were correct, there would be many cases of desires that would not bear agential authority because there would be no policies supporting identification or non-identification 166 i relations to them. This is so because from the fact that an agent has, for example, a con-policy concerning a desire it does not follow that she has a pro-policy concerning any other desire, and vice-versa. In addition, and what is even more problematic, there seem to be particular cases that involve identification with a desire even in the absence of a specific policy held by the agent concerning that desire.
Consider the case of Marianne who has a policy of not drinking more than two glasses of wine at dinner. Such a con-policy does not imply that she has a pro-policy towards having one, two or no glass of wine at dinner (not even that she has a policy of not drinking more than three glasses of wine at dinner).23 However, it obviously makes sense to ask whether Marianne is identified with her (potential) particular (token) desires for having one, two or no glass of wine at dinner. After all, if one of those desires occur in her psychology, she will need to ask herself whether to treat them as rea- son-giving. The question then is to know how the self-governing policy of not drinking more than two glasses of wine at dinner might help Marianne in such a case, and how exactly the policy is (rationally) related to her parti cular (token) desires of the moment.24 Assuming that increasing the number of policies to guarantee a one-to-one relation with every desire an agent might have is not a plausible option, Bratman owes us an answer to that.
On the face of it, Bratman seems to have two responses available. The first is simply to point out that self-governing policies are general and partial, and thus play only a general coordinating and organizing role in our prac- tical lives. This means that those policies play a role in constraining rationally the desires an agent might have by requiring a coherent network of rational relations (of consistency, coherence and stability) between the desires and the content of the policies. Thus, by having a policy of not drinking more than two glasses of wine at dinner, this policy is indirectly supporting the desire to drink one glass of wine at dinner. This desire could play, for example, an instrumental role in realizing that policy. This sounds correct, as far as it goes. However, the problem is that the sort of rationality constraint it appeals to is very different from the one involved in cases in which policies directly support a specific type of desire. This seems to alter significantly the relation between policies and desires.
The supporting relation of a policy towards a desire could now be interpreted in two ways. In Bratman’s basic sense, the supporting relation is one of determination. If someone has a policy concerning treating a specific desire as reason-giving, she has a commitment to treating it as end-setting on occasions in which it occurs. However, if the relation is extended to cases of desires which occur at particular moments but which are not di rectly supported by the policy-being only indirectly supported by it-the relation between the policy and the desire is different. Policies in those cases may play a regulative role, but not a determinative role as to how to treat a desire in practical reasoning. These are completely different forms of ra- tional constraint. A regulative role of a policy, by its own definition, does not specify completely what the agent is to treat as reason-giving on particular occasions. A regulative role is always incomplete in that respect. True, the regulative role played by a policy may confer or transmit rational justification to a desire that is compatible with it, thus “authorizing” an agent to treat the desire as reason-giving. But this cannot be a complete specification of the agent’s Identification with the desire. Surely, Marianne’s self-governing policy of not drinking more than two glasses of wine at dinner does not provide her with an answer to the question about identification with a desire for drinking one, two or no glass of wine (at a particular dinner time). This now means that there is something about identification with a desire which is left out and not exhausted by the role of self-governing policies.
The second answer available to Bratman is to draw on his notion of singular intentions (Bratman 2007d: 85 ff.). In spite of applying only to parti cular moments or being temporally restricted, singular intentions also carry rational demands of consistency, coherence and commitment to its exe- cution (though to a lesser degree in comparison to a self-governing policy, as singular intentions do not apply to a whole life nor to large parts of it). So, singular intentions could provide identification with a desire for particular moments in which an agent does not have a specific self-governing policy in favour of treating that type of desire as reason-giving. Insofar as the singular intention is compatible with the self-governing policy, Bratman might say that the agent is justified in adopting it.
However, this response will not do, for two reasons. Firstly, again, the previous problem persists: the higher-order policy might confer rational justification on the singular intentions compatible with it, but it does not provide identification with the desire which is the object of the singular in- tention. But, secondly, and more seriously, Bratman considers singular inten tions to be less representative of agential authority in virtue of their weaker connections to temporally extended agency. For Bratman, “the fact that such commitments [i.e. singular intentions] involve weaker connections to tem porally extended agency does, on the theory, entail that they have a weaker claim to agential authority” (Bratman 2007h: 189). This would now lead to the odd consequence that an agent’s being identified with a desire in virtue of its being appropriately connected with a singular intention, that is in turn compatible with a higher-order policy (that rationally “authorizes” the in tention), has “a weaker claim to agential authority”. It seems that something has gone awry here.
Actually, decisions, intentions and desires of particular moments might resist systematization in terms of self-governing higher-order policies and their role in cross-temporal agency. Decisions, intentions and desires of particular moments may be too a-systematic to get integrated into an overall perspective of cross-temporal agency and its general rational demands. So, the way the question for identification with particular decisions, intentions and desires is addressed and responded may be something that is neither given nor even guided by any higher-order self-governing policy. Still, they may bear the appropriate relations for being a genuine representative of the agent’s identification with them and, as such, for exhibiting no less agential authority than a higher-order self-governing policy. Humans are not only beings with a general view on practical matters, but also beings who live for the moment. They do not seem to be less autonomous for that.
Self-governing policies without identification
Human lives often take an unexpected course. Sometimes our choices turn out to be disappointing. Sometimes our ideals of a meaningful life get frustrated. We may end up feeling unfulfilled and thinking that life might be better, although there is nothing we can do to change it. Still, we can manage rationally such a life under the restrictive conditions imposed on us.
Those are quite common considerations and descriptions of how a life might turn out to be. However, as we will see, they create trouble for Bratman’s account of autonomy. This is bad news for Bratman, as a question about how a life goes on is surely relevant for his account, which presu- pposes an overall view of the agent’s life as embedded in a temporally ex tended sense of agency.
To make our point clearer, let us take, following Bernard Williams (1981b), the dramatic example of Gauguin. At some point in his life, Gauguin “turns away from definite and pressing human claims on him in order to live a life in which, as he supposes, he can pursue his art” (Williams 1981b: 22). He then abandons his family for a life dedicated to developing his ar- tistic skills, leaves Europe, becomes a solitary wanderer and finally settles in French Polynesia. As Williams hastens to add, the success of his choice “cannot be foreseen” (Williams 1981b: 23). So, let us suppose further that Gauguin’s choice of life got frustrated. To avoid running the risk of in- curring in historical ambiguities, let us call our counterfactual frustrated Gauguin “Cauguin”. So, Cauguin is an artistic failure (at least in comparison to his original expectations), he never gets used to a life outside Europe, he runs out of money, and loneliness makes him feel depressed. He ends up regretting and repenting of his choice, which he now describes as a mistake. Let us suppose also that there is no coming back for him (his family will not take him back, he has no financial resources for another change of life, etc.). Still, he keeps having the desires to lead an artistic life, to live in solitude, outside Europe, etc. Cauguin must then be able to manage his motivations and actions contextually under the practical restrictions he is now subject.
Let us now bring Bratman’s account into the picture. Where Cauguin now stands is the result of his having settled on a series of self-governing policies. In the face of no coming back, and despite his frustra- tions and regrets, he decides to reinforce his commitments to all his pre- vious self-governing policies-and perhaps come up with another series of self-governing policies to help him cope with his frustrations and regrets, his insistence on an artistic life of diminished value, his living poorly in so- litude, etc. He may then be able to rationally manage quite well his entire practical life from now on, and make sure that he sticks to all his commit ments in terms of consistency, coherence and stability. How autonomous is Cauguin?
For Bratman, at first sight, Cauguin would have all there is to say about being an autonomous agent. Still, there would be something missing and amiss in saying that, especially from the perspective of an account that finds in temporally extended agency the hallmark of autonomy. From this perspective, we would tend to refrain from saying that Cauguin’s life is fully autonomous. After all, he no longer identifies with it overall (and with its res pective motivational pieces and temporal parts), in spite of adopting self-go- verning policies that play a role in practically managing his life (in terms of consistency, coherence and stability over time). Cauguin’s case strongly sug- gests that it is one thing to be able to cross-temporally manage one’s practical life rationally (which may get justification for pragmatic reasons), and quite another thing to be identified with such a life (and with the respective mo tivational elements that compose it). Therefore, if this holds, self-governing policies do not guarantee identification.
Bratman’s response to cases like that would apparently come in two forms. The first would consist in simply reaffirming that Cauguin is fully au tonomous (Bratman 1999a: 199-200). In saying this, Bratman would seem to think that despite the fact that Cauguin’s decisions are not being made under optimal or ideal conditions, they are even so sufficient to lead Cauguin to settle on self-governing policies. Assuming that he is satisfied with them (due to absence of conflict among his self-governing policies), he is as auto- nomous as he can be.
This first response would simply miss the point. In Cauguin’s case, although he is not under optimal or ideal decision conditions, he certainly no longer identifies with his desires for leading an artistic life, living in solitude, etc. Thus, whatever policy Cauguin may hold concerning those desires, it cannot be a policy that justifies treating those desires as reason-giving as an end in themselves.
The second response is more interesting since it would grant that Cauguin is no longer identified with his desires for an artistic life, living in solitude, and so on, but claim that he may have self-governing policies for treating them as reason-giving in an instrumental sense (Bratman 2007h: 190-191). In other words, given that Cauguin still faces a self-management problem concerning his persisting motivations, but no longer identifies with them, he would do better to accommodate those desires into further policies by conferring instrumental justification on them. For example, po licies for treating as reason-giving the general desire for maximizing we- ll-being in the face of adversities or the desire for survival might both do the job.
However, Bratman’s second response also sounds unsatisfactory. First of all, we should notice that accommodating a desire with which the agent is not identified into the agent’s normative commitments by ascribing to it an instrumental role is not a paradigmatic case of instrumental rationality. Paradigmatically, an instrumental means is given justification by its being appropriately tied to a further and primary end. So, it is the achievement of the end that confers intelligibility to the means. But, according to that res ponse by Bratman, Cauguin’s desires would be given an instrumental role by the necessity of finding room for them among his normative commitments. So, their instrumental role would be arbitrary in that they are not primarily justified by any further end. This would make finding such an instrumental role for a desire an aim of practical reasoning, which may be acceptable as far as it goes. Surely, self-management may require on some occasions making non-paradigmatic uses of instrumental practical reasoning. However, the problem is to make it a case of exercise of autonomy. Let us develop this point further.
That response would mean that, at the end of the day, Bratman is trying to find normative space for an estranged element in one’s (normative) psychology. This sounds particularly odd for an account of autonomy like Bratman’s for which there is no way of assessing autonomy outside the temporally extended sense of agency, and which is intermingled with the agent’s personal identity. Cases like Cauguin’s suggest that there is more to agents’ identities and their own sense of agency as temporally extended than being capable of rationally managing their lives in terms of consistency, coherence and stability. Actually, the case strongly suggests that settling on self-governing policies might be a poor indication of an agent’s autonomy in its full sense. Due to Bratman’s emphasis on a temporally extended sense of agency, the question about autonomy should inevitably be also a question about the agent’s understanding her life as meaningful from her own perspective and, thus, connected with self-realization. But the relevant sense of “meaningful life” here does not seem to be captured essentially by rational demands of consistency, coherence and stability as they are embedded in self-governing policies playing a self-management role in one’s practical life.25
Bratman might insist that his account of autonomy severs the agents ideals and values from the self-managing picture of the agent.26 But this answer sounds awkward. The problem is not that an agent might consciously lead a life that falls short of her ideals and values. This is obviously possible. Granted, we would need an explanation of how the agent would manage her life in the face of it. And there is nothing amiss with finding here a candidate for an adequate sense of autonomy. But for an account like Brat- man’s this cannot be an exhaustive answer about what autonomy is. For an account of autonomy based on a temporally extended sense of agency and which is
intermingled with the agents personal identity, self-management as involving rational demands of consistency, coherence and stability over time can only be, at best, part of the answer about what autonomy is. If au- tonomous action requires an agents overall self-understanding of herself as temporally extended and as having a certain type of identity, autonomy in its fullest sense would bear a close connection with self-realization and with the agents leading a meaningful life. So, there would be no way of answering a question about a full sense of autonomy without including the ideals and values of the agent. They are part of her own identity over time. They are part of what really makes sense for her in leading her life.27
It is surely true that most of our lives fall short of fully realizing our ideals and values. Our decisions are always constrained by contingent conditions over which we have only limited control. This would make a conception of autonomy as a form of self-realization a matter of degree. We would be more or less autonomous depending on the extent to which we realize our ideals and values. But an account of autonomy understood as involving essentially a temporally extended sense of agency should welcome and embrace that implication, contrary to Bratman’s suggestion.