Abstract
In this paper, I propose a novel deliberation-based theory of autonomy which grounds an agent’s autonomy in her nature as a rationally-reflective being. I defend that theory against competing approaches to autonomous agency by arguing that the theory I propose is best equipped to handle two of the more troublesome problems that theories of autonomy face: the regress problem and the problem of manipulation. Sarah Buss and Peter Railton have each recently claimed that the regress problem which plagues many prominent accounts of autonomy indicates the need to abandon the notion that autonomous agency is to be understood in terms of deliberation, endorsement, or any other activity on the part of the agent, and adopt instead a “passive” approach to autonomous agency. Against this claim, I argue that despite their shift to the passive mode, the theories offered by Buss and Railton also face a version of the regress problem, and that the general solution to that problem implicit in their passive theories is available also to “active” theories of autonomy, such as the deliberation-based theory I propose. I go on to explain that because the solution to the regress problem I extract from the theories of Buss and Railton requires an “unmoved mover” of autonomy, the history of which is necessarily irrelevant, it invites manipulation objections. I argue that the theory I propose offers the most promising response to such objections, and thus escapes these two prominent problems in better stead than do the competing approaches to autonomous agency.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
This is of course not the sense of “autonomy” that all theorists aim to capture. Some theorists develop their conceptions of autonomy with an eye on other of the normative contexts in which such conceptions are put to work (e.g. Arneson 1994, who makes no mention of accountability but aims to develop an account of autonomous preferences with which to supplement preference-satisfaction conceptions of well-being) and others analyze personal autonomy without reference to specific normative issues (e.g. Mele 2001). While I believe that the account I propose could fruitfully be applied to several other normative contexts, I will in this paper focus exclusively on accountability. I will not here attempt to defend, but merely assume, the understanding of the relation between autonomy and accountability described by Buss, on which the former is necessary but not sufficient for the latter. I take for granted that there is indeed “a special sense in which agents determine their own actions when they are accountable for what they do” (Buss 2012, p. 648); my concern lies with the further task of identifying the relevant sense of self-determination.
For a thorough defense of this approach, see Mele (2001). The ellipses exclude “(or could have been).” As explained in footnote 13, some deliberation theories require only that an intention in some sense could have been produced by reflection. I set such theories aside here for reasons that are explained in that footnote.
Influential examples of this approach can be found in Dworkin (1970) and Frankfurt (1971). Taylor (2009) employs a modified version of this approach according to which autonomy requires, not that an agent endorse her actions, but instead that she endorse the decision-making procedures from which her actions result. As explained in footnotes 15 and 27, what I have to say about the more straightforward versions of the endorsement approach applies also to Taylor’s theory.
As Buss (2012, pp. 653–654) explains: “On an attitude-based conception of autonomous agency … the agent herself need not do anything to bring it about that she has the attitudes with which she is ‘identified.’ She need not engage in practical reasoning; nor need she weigh the considerations in favor of alternative attitudes.” While it is therefore possible to have endorsement without deliberation, there is a sense in which deliberation of the relevant sort will necessarily bring with it some form of endorsement. This is because the deliberation in question is that which results in an intention to act and, as Buss argues in a passage quoted in footnote 7 below, one cannot form such an intention without endorsing one’s action and the goal one seeks to achieve by means of it. There is thus a sense in which endorsement is necessary for autonomy on the deliberation approach, but only because this minimal form of endorsement is necessary for intentional action and the deliberation approach aims to answer the question of what more is required for an intentional action to also be autonomous. As explained in note 7, the fact that this form of endorsement is inherent to intentional action drives endorsement theories to employ a more robust conception of endorsement.
Taylor puts the problem in terms of an agent’s autonomy with respect to a desire, but it applies equally well to intentions and actions.
Buss’s rejection of the deliberation and endorsement approaches does not rest entirely on this regress argument; she levels additional objections against both approaches. In addition to its problem with regress, Buss argues that the endorsement approach will either set the bar for autonomous action too low by reducing it to intentional action, or else set that bar too high by implying that we act autonomously only when we act without reservation. If autonomous action requires only that the agent endorses her motives and/or actions—or, even less stringently, that she is not “dissatisfied” with those motives (Frankfurt 1999)—then it is effectively reduced to intentional action. This is because “one cannot make the sort of commitment essential to forming an intention without endorsing the goal one intends to achieve and without endorsing the intentional action itself insofar it is a means to this end” (Buss 2012, p. 654). Endorsement theorists might avoid this implication by requiring that the agent willingly endorses her motives and/or actions. This will not do, however, because autonomous agency is compatible with ambivalence: “without any impediment to our capacity for self-government, we can determine that the best thing to do under the circumstances is something we sorely regret having to do,” and something that we therefore do not willingly endorse (Buss 2012, p. 655). Her further objections to the deliberation approach are discussed in section 3, and summarized in footnote 18.
While Buss is, like myself, explicitly and exclusively concerned with autonomy of the sort necessary for accountability, Railton addresses both rational and autonomous agency, and he does not apply his theory to the issue of accountability. In this paper, I am interested in his only as a theory of autonomous, and not of rational, agency, and will below examine the implications that theory would yield for the issue of accountability, which Railton does not himself discuss. My aim there will thus be to examine whether Railton’s theory could serve as a plausible account of autonomy of the sort necessary for accountability.
See Buss (2012, pp. 661, 670, and 681).
She goes on to explain that “a trait is disabling in the sense relevant to our assessment of autonomous agency if, when it is a stable disposition, it typically prevents the members of an agent’s species from satisfying one or more of their basic needs without exceptional effort” (Buss 2012, p. 672). While Buss’s focus is on the autonomy of intentional actions, I presume that the intentions prompting such actions are also autonomous. It would be odd, after all, and also invite the ab initio objection, if an agent’s having governed the process by which she acquired the intention to act as she does conferred autonomy on her action but not on the intention that connects that action to the source of its autonomy. Buss (2012, p. 657) seems to acknowledge this, noting at one point that when agents are more fully involved in the production of their intentions, they are accountable not only for actions produced by those intentions, but the intentions as well. I assume that this is because the intention can also be regarded as autonomous.
See Buss (2012, p. 661).
As explained in section 4.4, it is open to Railton to hold that values, etc. can possess autonomy-conferring power, but only if they receive that power from the relevant capacities, which would thus remain the ultimate source of autonomy.
It is only versions of the deliberation approach requiring actual reflection that can and should employ this response to the regress/ab initio/incompleteness problem. As Buss notes, some deliberation theories focus not on whether an intention was produced by reflection, but instead on whether it could have been, on whether the agent had an opportunity to apply her reflective capacities to it. An opportunity of this sort is necessary for autonomy on the theory Mele offers in Autonomous Agents. Mele does not claim that such an opportunity is sufficient for autonomy, and rightly so, since to do so would be to ascribe unconferred autonomy-conferring power, not to any aspect of the agent’s identity or activity she has performed, but rather to opportunities for reflection. A theory of this sort would avoid the regress/ab initio/incompleteness problem entirely, but only by departing from the root idea of autonomy as self-determination. It is easy to see how capacities with which an agent can be identified could confer autonomy on the intentions and actions they produce; it is less clear how the mere opportunity to employ such capacities, which is not internal to the agent’s identity in any plausible sense, could do so. It is for this reason that I limit my attention in this paper to versions of the deliberation approach requiring the actual exercise of the relevant capacities.
This approach differs from Railton’s in that deliberative agency is, as he argues, only one domain in which humans are capable of responding aptly to reasons. As is discussed in greater detail in section 4.4, while adult humans are typically both reasons-responsive beings and rationally-reflective beings, these are distinct aspects of their identity.
It is useful to compare here Taylor’s modified endorsement account. According to that account, the degree to which a person is autonomous with respect to a decision and an action that decision prompts is determined by the degree to which the decision is the result of a decision-making procedure with which the agent is satisfied, where an agent “will be satisfied with a decision-making procedure if she believes that she has sufficient reason to continue using it” (2009, pp. 7–8). Hence, if an agent’s deliberative capacities possess autonomy-conferring power on Taylor’s account, this will not be because those capacities are internal to her identity as a rationally-reflective being, but rather because the agent is satisfied with them. If an agent is not satisfied with her deliberative capacities in the relevant sense, then they do not possess autonomy-conferring power, and if she is fully satisfied with an irrational decision-making procedure, then the decisions and actions in which that procedure results would be fully autonomous. As there are no further conditions that an agent’s satisfaction with a decision-making procedure must fulfill, it is with that satisfaction that autonomy-conferring power ultimately resides on Taylor’s theory. He does not discuss the regress/ab initio/incompleteness problem as it applies to the satisfaction that therefore lies at the foundation of his account, but I would suggest that Taylor is best understood as assuming a response to this trilemma akin to that I suggested more straightforward endorsement accounts could employ: as holding that an agent’s satisfaction with a decision-making procedure is autonomous, not because it is a product of some procedure with which the agent is satisfied, but rather because it is internal to her identity as a self-evaluative being. What is distinct about Taylor’s endorsement account would on this reading be that the primary object of the relevant self-evaluations would not be an agent’s motives and actions, but her decision-making procedures. As I explain in note 27, Taylor’s account nevertheless shares with more straightforward endorsement theories a vulnerability to the problem of manipulation.
There is thus a sense in which all four accounts are passive: an individual need not do anything to make these foundational sources of autonomy autonomous. But this is not the same sense in which Buss and Railton use the term. Their “passive self-determination” consists of a “doing”—the production of an intention on Buss’s account, the “embracing” of a reason on Railton’s—but a non-reflective “doing.” As an anonymous reviewer has helpfully pointed out, the contrast is thus between those theories that limit passivity to the foundational level and those that allow for a more thoroughgoing passivity by acknowledging this latter sense of “passive self-determination.”
Buss (2012, p. 659, footnote 29) acknowledges at one point that, by identifying different shared identities, other accounts of autonomy might join her own in holding that individuals act autonomously “only insofar as their intentions express an aspect of themselves they share with every other mature human being.” and she mentions their rationality as one such aspect. However, she fails to recognize that this offers any such account the same general route out of the regress(ab initio/incompleteness) problem that she takes herself, and that that problem therefore does not indicate that autonomous self-determination must be in the passive mode.
In total, Buss levels four objections against the deliberation approach. The regress problem and this over-demandingness objection are in my view the two most troublesome. The others are (A) that “since everything we do has a causal history … a person can fail to be accountable for her own deliberative activity” and (B) that deliberation theories lack a principled basis on which to distinguish autonomy-compatible from autonomy-undermining influences on an agent’s intentions, but typically simply list autonomy-undermining factors, such as brainwashing and hypnosis (Buss 2012, p. 651). The latter of these objections does not apply to a theory taking the route out of the trilemma I have described. Such a theory will make explicit that certain of an agent’s capacities possess autonomy-conferring power in virtue of the fact that they are internal to her identity as a rationally-reflective being. Factors such as brainwashing and hypnosis are not part of her identity in this sense, which is why when such factors exercise a decisive causal influence on the formation of an agent’s intentions (e.g. when such factors are used to implant an intention, rather than capacities which are then used to produce an intention—an importantly different case, as explained in section 4.3), her actions are not self-determined, but directly determined by something that is external to her identity. As explained below, a deliberation theory of this sort also has at its disposal a promising response to (A)—namely, to deny that the causal history of an agent’s deliberative activity is relevant to its autonomy and autonomy-conferring power, and to point out that Buss’s theory analogously, but more problematically, implies that the causal history of an agent’s functioning-compatible conditions is irrelevant to their autonomy and autonomy-conferring power.
This is not to say that it is only non-reflective actions that might indirectly receive autonomy in this way. As explained in section 4.3, the autonomy of the pro-attitudes that prompt an action is relevant to the autonomy of actions that are produced by way of reflection as well as those that are not.
For an example of how such a threshold might be specified, see Christman (1991).
Because such arbitrarily-initiated deliberations exercise capacities that help to make her a rationally-reflective being, they would possess autonomy-conferring power on this approach. The problem such an agent faces is thus not that she never acts autonomously, but rather that many situations calling for reflection will not occur at whatever predetermined or randomly-determined times she happens to reflect, and many important parts of her life will thus not be autonomously conducted. Even worse, it could be that deployment of her deliberative capacities is triggered, not by the clock or the flip of a coin, but by an evidence-detection ability that is “reversed” such that she always judges it necessary to reflect when it actually is not, and vice versa.
As with its deliberative counterpart, an agent’s evidence-detection ability need not be perfect, but merely minimally adequate. So long as that ability renders the agent sufficiently responsive to evidence indicating the need to employ her capacity to deliberate to help make her a (minimally) rationally-reflective being, it is internal to her identity as such a being. Sufficiency would here be spelled out in terms of the range of evidence to which the agent is responsive.
According to Arneson (1994, p. 49), “to live an autonomous life an agent must decide on a plan of life through critical reflection and in the process of carrying it out, remain disposed to subject the plan to critical review if disturbing or unanticipated evidence indicates the need for such review.” Similarly, Blöser et al. (2010, p. 239) claim that the autonomy of a pro-attitude requires (A) that the pro-attitude did not come about in a way that bypassed the agent’s capacity for critical reflection and (B) the agent possesses the capacity to critically reflect on that pro-attitude in the light of new experiences.
Where evidence is available to an agent in the relevant sense “if either (A) it would have been reasonable to expect the agent to recognize the evidence as offering a reason to [reflect], had she been possessed of minimally adequate evidence-detection abilities, or (B) it would have been reasonable to expect her to recognize the evidence as such in light of her more than minimally adequate evidence-detection abilities” (Weimer 2014, p. 281).
For a more detailed discussion of this example, see Weimer (2014, pp. 279–280).
As explained below, while that action is at least somewhat autonomous in either case, it would be more autonomous if autonomy had previously been conferred upon the pro-attitude prompting it by a relevant capacity than it would if autonomy had not been conferred upon that pro-attitude.
This is no less true of the satisfaction with one’s decision-making procedures that fills that role on Taylor’s modified endorsement theory. If a mad scientist implants in his victim both an irrational decision-making procedure and a belief that she has sufficient reason to continue using that procedure (i.e. “satisfaction” with that procedure), the decisions in which that procedure results cannot plausibly be regarded as autonomous such that the agent is accountable the actions they prompt.
While my focus will be on interference that implants positive conditions like these, it is in my view no more plausible to think that an agent’s autonomy is promoted by interference that serves to “merely remove the impediments of pathological influences,” as Buss’s account implies (2012, p. 687). If against an individual’s will a mad scientist removes some “impediment” or “handicap” that she has (in a minimally rational manner) reflectively decided to maintain, that interference should be taken not to enable autonomous agency, but to undermine it.
And thus more autonomous than would be an intention that was not reflectively-produced, but directly implanted – a plausible implication, in my view.
Consider, for instance, Robert Noggle’s (2005, pp. 101–103) account, according to which manipulation is compatible with autonomy if used to bring about the attitudes, commitments, etc. that make up an individual’s first “coherent and stable self,” but not if used “to implant psychological elements into an existing self.” As Valdman (2013, p. 18) argues, it is implausible to think that anything of normative importance (e.g. accountability) could in this way turn on which “bundle of desires, dispositions, and commitments … happened to get there first.” On the account I have outlined, by contrast, the order in which an agent’s motivational states appear is irrelevant; what matters is whether the agent has applied adequate reflective/monitoring capacities to them.
References
Arneson, R. (1994). Autonomy and preference formation. In J. Coleman & A. Buchanan (Eds.), In harm’s way: Essays in honor of Joel Feinberg (pp. 42–75). Cambridge: Cambridge University Press.
Blöser, C., Schöpf, A., & Willaschek, M. (2010). Autonomy, experience, and reflection. On a neglected aspect of personal autonomy. Ethical Theory and Moral Practice, 13(3), 239–253.
Bratman, M. (2003). Autonomy and hierarchy. Social Philosophy and Policy, 20(2), 156–176.
Buss, S. (2012). Autonomous action: self-determination in the passive mode. Ethics, 122(4), 647–691.
Christman, J. (1991). Autonomy and personal history. Canadian Journal of Philosophy, 21(1), 1–24.
Dworkin, G. (1970). Acting freely. Noûs, 4(4), 367–383.
Frankfurt, H. (1971). Freedom of the will and the concept of a person. The Journal of Philosophy, 68(1), 5–20.
Frankfurt, H. (1999). The faintest passion. In Necessity, volition, and love (pp. 95–107). Cambridge: Cambridge University Press.
Korsgaard, C. (1996). The sources of normativity. Cambridge: Cambridge University Press.
Mele, A. (2001). Autonomous agents. Oxford: Oxford University Press.
Railton, P. (2009). Practical competence and fluent agency. In D. Sobel & S. Wall (Eds.), Reasons for action (pp. 81–115). Cambridge: Cambridge University Press.
Taylor, J. S. (2005). Introduction. In J. S. Taylor (Ed.), Personal autonomy (pp. 1–31). Cambridge: Cambridge University Press.
Taylor, J. S. (2009). Practical autonomy and bioethics. New York: Routledge.
Valdman, M. (2011). Autonomy, history, and the origins of our desires. Journal of Moral Philosophy, 8(3), 415–434.
Valdman, M. (2013). Autonomy: incoherent or unimportant?. Unpublished manuscript.
Weimer, S. (2014). Autonomy and history: how a desire becomes one’s own. Journal of Moral Philosophy, 11(3), 265–293.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Weimer, S. Autonomy, Regress, and Manipulation. Philosophia 42, 1141–1168 (2014). https://doi.org/10.1007/s11406-014-9552-8
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11406-014-9552-8