Introduction
Before presenting my own model, I review the arguments made thus far. In Chapter 1 I explained that compatibilists like Harry Frankfurt posit causal necessity for their free action models in order to avoid randomness. I then argued that by relying on a neo-Humean notion of self such models nevertheless spiral into equally problematic arbitrariness because they do not enable the agent to act on desire-independent reasons, nor do they enable her to self-modify without encountering a vicious regress. I additionally argued in the first chapter that by placing all of the causal power in the reasons themselves, Susan Wolfs approach, while it at first appears to render agent action rational, actually renders both rational and irrational action impossible. Genuine alternative possibilities are necessary to qualify the processes of reasons-recognition, acting on reasons and self-correcting as rational.
In the second chapter I examined Timothy O'Connor's theory of a rational capacity by agent causation, a theory whose underlying motivations are similar to my own. O'Connor attempts to build a model according to which the agent exercises a special, non-event-causal rational capacity, identifying a range of genuine alternative possibilities and then selecting one among them. Because O'Connor does not directly analyze the status or nature of the reasons on which such an agent can act and because he does not develop the sense in which she may use contrastive reasoning to select her actions, his model as it stands ultimately must succumb to the arbitrariness charge.
O'Connor does not tell us on what basis the agent makes her choices and how she comes
to have the criteria for selection that she does. Though he rightly stipulates a gap between the reasons and the action, he merely fills that gap with "agent causation," a form of causality that does not really reveal how a decision is freely made.
This chapter attempts to lay out a viable third alternative to event-causation and randomness, namely intrinsically unpredictable rational action. The past two chapters have revealed the extent to which the debate on free will has been shaped by a perception that any model for agent action will either encounter randomness or require causal necessity. This underlies one major concern that compatibilists have about libertarian models of free agency. It is the worry that if an agent does not perform an action for causally sufficient reasons, then she performs it for causally insufficient ones. Unless there is some additional element that brings the insufficiency to sufficiency without bringing necessitation along with it, it seems to the compatibilist that such an agent must be acting, at the final moment, randomly. If there is a gap between the reasons, which are not causally sufficient for any one action, and the action performed, the compatibilist wonders what could possibly fill it. I argue that if we deny the neo-Humean conception of the self and replace it with a unified, active, and transparent rational capacity, we have a way to fill the gap.
The approach here differs from the ones taken by libertarians like O'Connor and Clarke in that it takes the problem of the ontological status of desire-independent reasons for action head on, revealing how one can act on them without being causally necessitated by them. In this chapter I describe the capacity for reasons-recognition and acting according to reasons that is necessary for freedom of the will. In order to enable the agent to utilize this capacity freely, the capacity has three necessary features. The
first is that the agent must actively and consciously choose between alternatives. She must face genuine alternative possibilities and it must be the case that, were an observer to know everything possible about the agent, he would not be able to predict what she would use her capacity to choose next. Her choices would not simply reflect a passive manifesting of her given capacity. The capacity is not a structure; it is an active ability.
The second necessary feature of my model is that it accepts a unified notion of self.
Unlike the neo-Humean self, whch is constituted by various contingent psychological elements, each with its own motivational tendencies, the unified self on my view does not have various dispositional elements that are distinguishable from one another. It is not as though the agent happens to have a few contingent rational qualities. Rather, her rationality characterizes the conscious perspective according to which she navigates the world. The third feature of the rational capacity is its transparency. In order to be able to actively recognize and select among reasons, the agent must be aware of them. The neo- Humean notion that an agent may be caused to behave a certain way by a desire of which she is unaware is incompatible with the model for free action defended here.59
In order to show how the rational capacity defended here does not encounter arbitrariness in the way that 07Connor's generic agent did, I will argue that by normatively shaping themselves agents can individualize their capacities in rationally defensible ways. Such normative self-evaluation does not, as O'Connor argued, involve the development of various tendencies to act in certain ways." Rather, the agent
59 I do not mean to deny that the possibility that latent desires sometimes cause the agent to act. I mean instead, that if a latent desire so causes in a particular situation, then the agent has not acted freely. See appendix: objections.
60 "This conception of agent causation [is] as a structured capacity - structured by tendency-conferring states of having reasons to act in specific ways and more enduring states of character, involving relatively fixed dispositions and long-standing general intentions and purposes around which one's life has come to be organized," O'Connor (2000), pp. 97-98.
individualizes her rational self using a process of normative self-evaluation that operates, like the processes of recognition of reasons and acting according to them, with strong PAP throughout. An agent's normative self-conception relies on her development of internal reasons, which she recognizes by applying nonnative reasons to given features of herself as an individual. O'Connor's agent was the sole causal source of her deliberation and action selection. But because O'Connor did not describe how the generic agent's causal power could be operated by the agent as an individual, she had no way to non- arbitrarily choose between various rational options. The normative self-conception I develop here meets two potential objections. Navigating between various worthy reasons is not arbitrary on my model because normative self-evaluation provides the agent with an individualized but rational basis for choice selection. Second, agents are not constrained by their rationality because this individualization means that not all rational agents would respond the same way, given the same situation.
Section I. Non-causal Reasons
The eliminative effect that a causal conception of action has on the very possibility of rationality, the objective authority of reasons, the justifiability of actions, and a normatively structured conception of self, renders it an insufficient model for agent action. The previous chapters were meant to challenge the compatibilist to openly acknowledge the full extent of the reduction of these notions that a universal event-causal model would bring about. To consider reasons as causes is to give up on reasons because framing them this way disables them fiom offering justificatory explanations, which is their primary job.
An explanation of reasons as something other than mere causes, in contrast, affords the notions of rationality, justification and normative motivation a role in human action. It is the third and independent alternative to the traditional "causally necessitated"
or "random" dilemma. Because a crucial objection to the possibility of this model has typically centered around the notion that reasons, normatively understood, do not have motivating power, I will now sketch a reasons-explanation model with motivating elements built in. Though these reasons have objective weights, their weights are not wholly determinate and, as such, they do not provide causally sufficient conditions for just one action, given a particular situation. In addition, because my proposed model does not expect one contingent desire to be able to judge the worthiness of another, it does not lead to the infinite regress of selection criteria that plagued the compatibilist deep-self theorists. It avoids the infinite regress by enabling the agent to use normative criteria of appropriateness and worthiness to evaluate her reasons.
Objectively authoritative, desire-independent reasons include: external reasons, meaning various obligations to others; internal reasons, which include the individual's goals and ideals; and reasons that reflect aesthetic or pragmatic values. Desire- independent reasons of value and individual ideals will reflect considerations of worthiness, while those of external reasons or 'oughts' will reflect considerations of appropriateness. Naturally, as human agents are not perfect reasoners, reasons can be derived from false beliefs about the facts in the world, but an action based on such a reason can still be rational with respect to that false belief. Reasons can be anticipatory, and can anticipate consequences to greater and lesser degrees of accuracy, but the essential characteristic of desire-independent reasons for action is that they can provide
not only the full explanation, but also the justification for any given rational action. The inability on the agent's part to provide any justificatory reasons for his action qualifies that action as irrational in a way that psychologically contingent and necessitated reasons, which are always fully explained by their event causal status, cannot.
The claims on the part of compatibilists that all reasons must originate in desires and contain some desire-based motivation is false. The content of all reasons need not include volitional elements. Some desire-independent reasons favor particular actions on their own. The fact that a building is burning, for example, provides a good reason to leave it that has nothing to do with an agent's volitions. Such desire-independent reasons may not be appealing in the sense of desire-invoking and they may not be compelling in the sense of emotion-invoking. But they reflect objective rational standards that most of us within a certain threshold of Wolfian sanity are able to recognize and on which, even in the absence of a corresponding desire, we are able to act." Just as it is the case in forming a belief that one merely needs to judge that there is enough evidence to indicate that the belief is true, not that one have the desire to so believe, so it is with action as well.
If one recognizes a number of reasons that favor action x and no reasons that favor some other action," then this recognition is grounds enough to motivate the intention to No further desire or want to x need be assumed.
The connection that such reasons hold to the various actions available is, in many cases, a practical, instrumental, or theoretical relation. When O'Connor writes that an agent performed a particular act "in light o f ' a certain reason, I claim here that this "in
61 Recall Wolfs definition of sanity from Chapter 1.
62 "Some other action" includes the choice to perform no act at all.
63 Scanlon, T. M. What We Owe to Each Other, (1998), pp. 33-34. In a footnote Scanlon attributes this point especially to Christine Korsgaard in "Skepticism about Practical Reason."
light o f 7 relation is the normative relation that exists between the various facts and the specific act." The act chosen relates to a fact as a reason not in some mysterious way, but rather in the most commonsense way possible, much like the numbers two and four relate such that 2 + 4 = 6. The answer is six "in light o f 7 the relation between numbers two and four when they stand in the operation of addition. To act "in light of a reason" or to say that a particular reason "favors" a particular act is merely to recognize what is really the case, to solve a problem that presents itself and to do so accurately.65 The ability to conceive of decision-making situations this way relies on our possessing a well- functioning rational capacity, which will be described and defended in the next section.
As we see from the paragraphs above, reasons relate facts and potential acts both justificatorily and explanatorily. Though in much of the literature on reasons explanatory reasons are considered to be those that provide desire-dependent motivations for acts, here "explanatory reasons" includes those that provide a motivation for the action that can be defended as rationally justifiable. The reasons-as-causes model, on the other hand, is not able to judge the quality of the explanation for a particular action. The answer to the question "why that particular act?" is simply "the state of affairs at t,." Such an answer is not equipped to offer justification for the action.
Reasoning indicates the kind of evaluation and calculation that involves recognizing the degree of nonnative value of various elements within a given situation and of figuring out the optimal path. The optimal path will not only fulfill rational third- person criteria, but will also honor the normatively structured internal reasons the agent
64 O'Connor (1995b), p. 270.
65 "Reason statements are relational in three ways. First, the reason specified is a reason for something else. Nothng is a reason just by itself. Second, reasons for action are doubly relational in that they are reasons for an agent-selfto perform an action; and third, if they are to function in deliberation, the reasons must be known to the agent-self," Searle (2001), p. 99.
has conceived of for herself, including her long-term goals and personal c ~ m n l i t m e n t s . ~ ~ In other words, the agent does not necessarily pick the one path that rational criteria deem optimal, but picks according to her own reasons that she is able to assess according to third-person criteria. Her use of her rational capacity is therefore particular to her, but continues to enable her to defend her actions as being based on justificatory reasons.
This subtler navigation of the rational capacity is necessary for her to be able to select among several rational alternatives.
Section II. Recognizing Reasons and Acting According to Them
A rational capacity that allows agents the freedom to select among multiple options must have three features: activity, unity, and transparency. Only if the agent has a unified self that uses her rational capacity actively and with awareness of the reasons she is considering will the capacity successfblly fill the gap between the reasons there are and the agent's recognizing and acting on them. Only then will such a capacity provide a successful alternative to causal necessity and randomness.
In this section I claim that it is the agent herself, in her capacity as rational and sane evaluator, who actively fills the gap between the reasons and the action she brings about. It is a feature of such an agent that she is able to recognize, assess, and select among reasons that are not causes, but rather relations between facts and hypothetical actions. It seems to me that the push for agent-causation is meant as a stop-gap solution to the infinite regress worry. This is the worry that an agent's criteria of selection will be arbitrary and therefore require some freely chosen principle of selection, a principle that seems unavailable to the agent who is not able to self-create. Positing agent-causation
66 More on ths will be presented in Section 111.
implicitly accepts the compatibilist view of self, which emphasizes the agent's being capable of freely selecting whatever weights for reasons she feels like and which ignores the objective worthiness that I assert is detectable in the reasons themselves. The rational capacity view outlined here instead forthrightly recognizes the objective worthiness of certain values, the objective appropriateness of certain acts, and declares that we are the kinds of agents who are able to recognize such normative reasons for what they are. In this sense I return to Wolfs model in some ways, affirming her views on the importance of the sanity of the individual, her ability as a reasoner, for the possibility of free and rational action. In this section, along with the next two, I will distinguish my model from the two extremes of a rational, albeit generic and un-individualized capacity and an individualized, but not rational neo-Hurnean self.
The first and most essential feature of the rational capacity outlined here is its activeness. I have thus far denied that a coherent model of free-agent action can rely on any form of event-causality, primarily because I have argued that the assumption of necessitating causality immediately forces arbitrariness onto the model and makes genuine rationality impossible. If, I have argued, the free-agent-action model is event- causal, the notion of an agent's ability to act one way or another is unintelligible. But if the agent faces genuine alternative possibilities and is not caused to select one or another, then the capacity that enables her to make a decision based on the available reasons must be active. She must transform one out of a range of causally insufficient reasons into the reason on which she acts. If the reasons there are do not have the capacity to cause us to recognize them or act on them, then we must actively pursue them. Nothing else can
"make" us do so. The idea is that if we were to sit back and wait for ourselves to be
causally propelled to act, no rational acts would ever ensue, though non-rational behavior might.67
Essential to the fieedom-conferring rational capacity is the ability on the agent's part to step back and evaluate the reasons, to consider conflicts between them, their normative weights and the actions they would justify. The rational capacity is not pushed and pulled by the reasons, but instead actively, not merely perceptually, recognizes them, evaluates them according to their not-fully-determinate objective authority,68 and acts on them such that the reasons can provide the content of the agent's action, the motivation for her action, and can be presented as justification for her action, according to rational criteria. Most importantly, though, given that the reasons are causally insufficient, the process of rational recognition, deliberation, and intention-formation is necessary for both the explanation and justification of the action to be complete. The rational-capacity model contrasts categorically with the psychologically-triggered model and provides the agent with two abilities. First, it enables agents to recognize reasons that motivate their actions without causing them and second, it enables them to provide genuine normative justifications for their actions.
In this paragraph I contrast perceptual capacities, which are passive and therefore not freely exercised by the agent, with recognitional capacities, which are active, meaning imbued with strong PAP and under the free control of the agent. To freely act on a reason, one must know the facts relevant to the situation and be able to identify which actions one could pursue. The former is a perceptual capacity because one cannot
67 We would, of course, continue to breathe and blink and physically react to stimuli.
68 The objective reasons I posit are not fully determinate because they are non-causal.