1

III. The Costs of Hard Compatibilism

“There is much more to say about free will, and the point made in this essay is just the beginning. But it is the beginning. It is important to be clear about it, and to try not to avoid or occlude it in any way.”

–Galen Strawson[28]

Hard compatibilists are bashful about the costs of their view. This section will attempt to show just how unattractive some of the consequences of hard compatibilism are. Some of these costs are undeniable; others are more controversial but, this article will argue, nevertheless actual. Necessary costs include the inability to guarantee: that an agent’s behavior cannot be perfectly predicted, that an agent’s exercise of control is not itself subject to meta-control, that the agent is uniquely responsible for her or his actions, and that free will is an all-or-nothing phenomenon. Controversial costs include: an inability to explain why belief in free will is often subject to doubt and a loss of philosophical elegance.

To explore these costs, it will be helpful to focus upon a curiously steadfast presence in the free will literature: the meta-controller. A meta-controller is the controller of another controller (the controllee). Hard compatibilists and those who deny the existence of free will both have their own meta-controllers. Hard compatibilists (especially semi-compatibilists) appeal to the meta-controller in Frankfurt’s famous example.[29] Similarly, those who deny the existence of free will (especially hard incompatibilists) appeal to the “neuroscientists” in Pereboom’s famous four-step argument.[30] These arguments show how meta-controllers might control their victim earlier in the past by programming the agent to perform specified actions in the future. Indeed, one can imagine how God—the quintessential meta-controller—programs our entire lives.

Necessary Costs

First, consider a series of meta-controllers. There might be N number of meta-controllers. We may label these meta-controllers MC1, MC2, and so on, in chronological order. Each meta-controller need not know of the existence of its own meta-controllers. In this way, we have a series:

Initial Conditions à MC1 à MC2 à … à Controllee

This picture raises an important question: what responsibility do these controllers have with respect to each other’s behavior?[31] The author suspects that there are only two answers to this question: all are responsible or none are. Hard compatibilists adopt the former position; those who deny the existence of free will adopt the latter. Exploring the relationships between these controllers illustrates the costs of hard compatibilism.

Predictability

The first cost of hard compatibilism one might notice is that agents can no longer guarantee that an agent’s life is not absolutely predictable.[32] Certainly, human behavior is to a large extent predictable. One can predict, for example, that most of the world’s population will fall asleep at some point during the next twenty-four hours. Yet one resists the notion that one’s life is absolutely predictable. Perhaps I will go to sleep in the next twenty-four hours, but I will decide exactly when that event takes place. One would like some “elbow room”[33] in our lives. Without this “elbow room,” the agent would not seem to be making any meaningful contribution to its life. But this is precisely what hard compatibilism cannot guarantee. Hard compatibilism entails the possibility that a meta-controller, given sufficient information about the universe, could predict one’s entire life. This is so because hard compatibilism, by definition, further admits that a meta-controller might design one’s entire life.

Meta-Control

If a meta-controller M1 might have designed one’s entire life, then M1 does seem to exercise a sort of distant control over one. The meta-controller has this control because it can determine how one lives one’s life according to its own, and not the agent’s, prior wishes. For example, if M1 wishes for the controllee to be a woman who decides to marry on her 21^st birthday, then one will be a woman who marries on her 21^st birthday.

Hard compatibilists argue that there are differences between this sort of control and other exercises of control.[34] Of course there are. To satisfy the critics, one might refer to the control that a meta-controller exercises over a controllee as control*. The control that a kidnapper exercises over a hostage is different from the control that a meta-controller exercises over what-becomes-an-agent: a hostage has preexisting wishes that are violated whereas what-becomes-an-agent has no preexisting wishes at all. The question, however, is whether these differences are relevant to evaluating the costs of hard compatibilism.

I suggest the answer is that these differences between control and control* are not relevant. The differences are not relevant because both types of control possess a feature that one might, according to prephilosophical intuitions, find disturbing: foreignness. In both cases, whether the agent has preexisting wishes or not, what determines the whole story of the agent’s life is not the agent itself. In this respect, the agent at least seems to have less control over its own life.

These examples of meta-controllers may be surprising. One might, according to pre-philosophical intuitions, consider control to be something incompatible with meta-control. If someone else controls how I exert control, then do I have genuine control? One might regard the control that a controllee possesses to be a sham. Bishop Bramhall famously thundered against the arguments of Thomas Hobbes for just this reason:

“I hate this doctrine from my heart... It destroys liberty, and dishonours the nature of man. It makes the second causes and outward objects to be the rackets, and men to be but the tennis-balls of destiny. It makes the first cause, that is, God Almighty, to be the introducer of all evil and sin into the world... Excuse me if I hate this doctrine with a perfect hatred, which is so dishonourable both to God and man; which makes men to blaspheme of necessity, to steal of necessity, to be hanged of necessity, and to be damned of necessity... It were better to be an atheist, to believe no God; or to be a Manichee, to believe two Gods, a God of good and a God of evil; or with the heathens, to believe thirty thousand Gods: than thus to charge the true God to be the proper cause and the true author all the sins and evils which are in the world.”[35]

Similarly, Hume found the costs of hard compatibilism to be so great as to conclude that philosophy has yet to justify them and should instead focus upon “common life”:

“To reconcile the indifference and contingency of human actions with prescience; or to defend absolute decrees, and yet free the Deity from being the author of sin, has been found hitherto to exceed all the power of philosophy. Happy, if she be thence sensible of her temerity, when she pries into these sublime mysteries; and leaving a scene so full of obscurities and perplexities, return, with suitable modesty, to her true and proper province, the examination of common life; where she will find difficulties enough to employ her enquiries, without launching into so boundless an ocean of doubt, uncertainty, and contradiction!”[36]

Unique Responsibility

A third cost of hard compatibilism is the uniqueness of responsibility. If a meta-controllers control controllees then, even if the controllees can obtain some responsibility for their actions, they must share this responsibility with their meta-controllers. Moral responsibility multiplies.[37] Hard compatibilism cannot guarantee that agents alone are morally responsible for their lives.

In ordinary responsibility practices two agents may share responsibility for a given event. For example, if two people simultaneously, and independently, shoot rifles at an innocent victim, then they can share responsibility for the victim’s death. But, although agents might share responsibility for proximate event, they usually consider themselves uniquely responsible for immediate events. So each shooter might consider herself or himself uniquely responsible for the event “pulling the trigger.” According to hard determinism, however, the shooters could not even claim this much. They must share responsibility for their entire lives.

All-or-nothing

A fourth and final necessary cost of hard compatibilism is its failure to regard free will as an all-or-nothing phenomenon. Hard compatibilists consider this a virtue because, in ordinary practice, moral responsibility does admit of degrees. The concept of free will, however, has traditionally been regarded as all-or-nothing.[38] Although, in ordinary practice, moral responsibility of admits of degrees, ordinary responsibility practices also neglect the philosophical dilemma of free will. That philosophical concept is associated with more extraordinary metaphysical and religious claims. Free will is, on this view, like a soul: indivisible. By abandoning free will’s traditional attributes, such as its being all-or-nothing, hard compatibilism takes on a revisionist character.

Control**

The hard compatibilist may not feel impressed by these arguments. Perhaps such a hard compatibilist is an atheist who remains convinced that life evolved on earth and no designer created it. The prospect of a meta-controller exercising control over one’s life would not be especially disturbing, not because the prospect itself fails to disturb, but because the prospect is not actual. But this hard compatibilist’s relaxation is too quick.

Hard compatibilism, by definition, asserts that manipulated agents, if not improperly manipulated, may still possess free will. This symmetry between manipulated and non-manipulated agents should hold, however, for hard compatibilism’s costs as well as its benefits. Indeed, one might argue that initial conditions can exercise control* over an agent just as well as a meta-controller can. Again, initial conditions have the same essential feature that meta-controllers do: foreignness. An agent feels no particular ownership for these initial conditions, yet they determine the story of her or his life. In the same way, whether one was pushed down by pranksters or hurricane winds, in both cases one is not morally responsible for falling. Both pranskers and hurricane winds are foreign to the self. One can refer to this latter sort of control as control**. Hard compatibilists should find control** to be just as disturbing as meta-control.

Controversial Costs

The previous costs are uncontroversial and hard compatibilists admit them—even if they do not always call attention to them. The following costs, however, are more controversial. I will argue, however, that they are nevertheless actual.

Error theory of PAP

The greatest cost of hard compatibilism is that it fails to explain why anyone would have ever found the existence of free will controversial. Free will, according to hard compatibilist definitions, is not just unproblematic, but also obviously unproblematic. No one ever doubted, for example, that ordinary people are moderately reason-responsive or have lower level desires in accordance with their higher level desires.[39] Furthermore, hard compatibilist views fail to explain why anyone would demand more, for free will, than these views offer. Hard compatabilists lack an error theory for incompatibilism. By describing the costs of hard compatibilism, this article helps to fill that gap.

Indeed, those who deny the existence of free will can provide their own error theory of libertarianism. Libertarianism defends the Principle of Alternative Possibilities (PAP), according to which moral responsibility requires alternative possibilities. Those who deny the existence of free will can offer the following error theory of libertarianism: PAP represents the defender of free will’s misguided, and futile, attempt to escape the costs of hard compatibilism by asserting that people are causa-sui.

In contrast to hard compatibilists, defenders of free will denial usually endorse a concept of free will that is all-or-nothing: being a self-cause or “causa-sui.”[40] Being cause-sui, in this sense, involves having nothing foreign to the self contribute to its own creation. Rather, the self creates the self in a vacuum—ex-nihilo.[41] This is, of course, logically impossible. But it avoids the costs of hard compatibilism. An agent who is causa-sui would be unpredictable, could not be the victim of meta-control, could not share moral responsibility with anyone, and would possess an all-or-nothing type of free will.

Of course, AP alone will not suffice to make one causa-sui. But the insufficiency of AP, in this respect, is less obvious than the insufficiency of its denial. In a deterministic world, one can easily trace aspects of one’s self to those obviously beyond one’s control. In an indetermistic world, however, one can only trace these attributes to branching points in the universe. A little philosophical rigor shows that one cannot attribute how these indetermistic events unfold to anyone. That is the essence of indeterminism. The libertarian, however, cannot resist the temptation to attribute these events to the self. This is ironic considering that the entire purpose of PAP was to deny the relevance of antecedent factors, over which the agent has no control, to that agent’s choices. What the libertarian forgets is that the self belongs to that set of antecedent conditions.

Once one recognizes the importance of being causa-sui, one also recognizes a time shift within the free will debate. Traditionally free will has been conceived of as just an immediate or “time-slice” power. But what these considerations show is that the problem which motivated the free will dispute is historical. PAP attempts to compensate for the fixity of the past by denying the fixity of the present. But once the libertarian grants that a given self is problematic, there is no self remaining to exploit these current alternative possibilities. Agents cannot create themselves ex-nihilo. From this perspective, one recognizes the charge that free will does not exist, if such a charge is to capture the genuine dilemma which motivated the dispute, is not a claim about immediate or “time-slice” powers. Rather, the charge is that one’s entire life is ultimately arbitrary. Everything, on this view, is an accident of birth. Because one’s life is ultimately arbitrary, one cannot claim responsibility for it. Daniel Dennett ably described what the libertarian wants:

“He wants above all for the decision to be ‘up to you,’ but if the decision is undetermined—the defining requirement of libertarianism—it isn't determined by you, whatever you are, because it isn't determined by anything. Whatever you are, you can't influence the undetermined event—the whole point of quantum indeterminacy is that such quantum events are not influenced by anything—so you will somehow have to co-opt it or join forces with it, putting it to use in some way, an object trouvé that you meaningfully incorporate into your decision-making in some fashion. But in order to do this, there has to be more to you than just some mathematical point; you have to be someone; you have to have parts—memories, plans, beliefs, and desires—that you've acquired along the way. And then all those causal influences from the past, from outside, come crowding back in, contaminating the workshop, preempting your creativity, usurping control of your decision-making. A serious quandary.”[42]

The word which illustrates Dennett’s genius is “contaminating.” Incompatibilists feel infected by arbitrariness. In order to escape this arbitrariness, libertarians deny the importance of foreign contributions to the self. They continue to shed this self, infected by arbitrariness, until there is nothing left but a “mathematical point.” Being more than a mathematical point, however, is a prerequisite of being an agent.

One might illustrate the dilemma with the following analogy. Saul Smilansky says that we are all the “unfolding of the given.”[43] In this sense, one might imagine a person, like a child at Christmas, receiving a gift. The gift in this analogy, however, is the person’s self. This self will include their heredity and environment and everything else about them that they did not choose. According to those who deny the existence of free will, those who believe in free will commit a sort of false self-ownership. The compatibilist receives this gift and cherishes it without critical reflection. Similarly, libertarians feel that they receive this gift of self with approval only after critical reflection. Perhaps they think they have given this gift to themselves. But receiving one’s self as a gift is, of course, logically impossible. In reality, there is no self yet to receive this gift. So the libertarian picture is false. The compatibilist might understand the situation but remains unmoved. As John Fischer has written: “I seek to call attention to the fact that luck (of the relevant sort) is pervasive and, upon reflection, not problematic (in any context).”[44] Only those who deny the existence of free will recognize how disturbing this situation: if there is no self yet to receive this gift, there can be no self to critically examine whether this gift should be received. The gift itself determines whether it will be accepted. It is impossible for any given agent to conclude that the way the agent is is the way the agent ought to be.[45]

This error theory of libertarianism also captures what the libertarian means by the phrase “up to” one. As Dennett notes, libertarians want their decisions to be “up to” them. The precise meaning of this phrase, however, remains obscure. This error theory of libertarianism suggests an immediate meaning, however: a decision’s being “up to” one precludes the decision from being available to another. So a decision’s being “up to” one precludes another from reliably predicting one’s decision before one decides. This other’s ability to reliably predict one’s decision implies that the decision was not genuinely “up to” one, but rather “up to” some third party, which the other can observe. Similarly, a decision’s being “up to” one precludes a meta-controller from exercising meta-controller over one, because the decision is not “up to” the meta-controller. In the same way, a decision’s being “up to” one precludes a meta-controller from sharing responsibility for a decision. To share in responsibility for the decision, the meta-controller would need the decision to be in some way “up to” the meta-controller. But libertarian decisions cannot be “up to” the meta-controller; they are “up to” the agent—just the agent. An agent’s freely willing that X precludes any other agent from ever freely willing that X. The choice is the agent’s alone. The free will that libertarians want is exclusive.[46]

The libertarian desire for exclusive free will helps explains the common complaint that compatibilist versions of certain goods are shams. These goods include creativity, autonomy, desert, moral responsibility, reactive attitudes, dignity, individuality, life-hopes and love.[47] One might regard the compatibilist version of each as a sham because one may attribute these compatibilist versions, not just to the subject, but to other agents (or non-agents). For example, the reactive attitudes involve reactions towards agents (or non-agents) with respect to certain actions. If someone punches me in the face for no apparent reason, I might feel the reactive attitude of anger. But if God designed that person’s entire life such that the person would punch me in the face, then I must share my anger between the person and God. Similarly, if God designed my entire life such that I would fall in love with a woman, then my love seems less pure to the extent that it arises from at least two sources: myself and God. The libertarian chases a love that springs, ex-nihilo, from the agent itself. Such love is impossible.[48]

This error theory of libertarianism, according to which libertarians seek to avoid the costs of hard compatibilism, also captures what Nietzsche meant when describing “‘freedom of the will’ in the superlative metaphysical sense”:

“The causa sui is the best self-contradiction that has been conceived so far, it is a sort of rape and perversion of logic; but the extravagant pride of man has managed to entangle itself profoundly and frightfully with just this nonsense. The desire for ‘freedom of the will’ in the superlative metaphysical sense, which still holds sway, unfortunately, in the minds of the half-educated; the desire to bear the entire and ultimate responsibility for one's actions oneself, and to absolve God, the world, ancestors, chance, and society involves nothing less than to be precisely this causa sui and, with more than Munchhausen's audacity, to pull oneself up into existence by the hair, out of the swamps of nothingness.”[49]

Being causa sui, Nietzsche suggests, is incompatible with predictability, meta-control, and sharing responsibility (“God”). Similarly, it is incompatible with meta-control** (“chance”) and considers foreign contributions to the self to be disturbing. So libertarians try to shed each of these foreign contributions (“God, the world, ancestors, chance, and society”) until they are nothing but a “mathematical point” trying to pull themselves “out of the swamps of nothingness.” By acknowledging the real costs of hard compatibilism, free will denial can provide this error theory of libertarianism that is unavailable to the hard compatibilist: the libertarian project begins by evading the costs of hard compatibilism and ends in suicide.

Philosophical Elegance

Nietzsche’s aphorism also captures another essential characteristic of the free will dilemma: its philosophical elegance. Nietzsche identifies the free will dilemma as a bootstrapping problem.[50] Indeed, he describes the problem as trying to pull one’s self up “into existence”, not by one’s bootstraps, but “by the hair, out of the swamps of nothingness.” A bootstrapping problem is one in which any solution must presuppose or rely upon assumptions which are themselves problematic. Such solutions are self-undermining. The relevant assumption within the free will problem is the assumption that control presupposes freedom from meta-control or the assumption that an agent may already exist and choose which wishes or preferences the agent should have. But this characterization immediately presents a vicious regress. Controllers will only exercise control in accordance with meta-control* or meta-control**. Having wishes or preferences is a prerequisite of being an agent. So no agent can solve this vicious regress.

Although this characterization renders free will impossible, it has the advantage of being philosophically elegant. It is not a coincidence that Galen Strawson considered his Basic Argument, which argues against the existence of free will, so trivial and obvious that he discusses it in just one chapter of his book Freedom and Belief.[51] In contrast, hard compatibilists such as John Fischer and Daniel Dennett make much longer defenses of their respective views.[52] This is not a coincidence because Strawson’s argument is a priori and examines an all-or-nothing conception of free will. Fischer and Dennett, however, examine conceptions of free will which are not all-or-nothing. This forces them to consider boundary cases. In summary, Strawson’s conception of free will is decidedly simpler, and more elegant, than hard compatibilist alternatives. In contrast, Strawson’s conception presents precisely the sort of dilemma which an ancient Greek philosopher might have immediately grasped and confronted.

Charactering the free will problem as a bootstrapping problem also enhances its philosophical elegance by preserving its similarity to other philosophical dilemmas.[53] I am not an expert in epistemology or ethics but the fundamental problems of these subjects strike me as bootstrapping problems as well. For example, in epistemology, any assumptions about knowledge (such as the alleged fact that other people exist) depends upon further unwarranted assumptions about knowledge (such as the alleged fact that one is not a brain-in-a-vat).[54] Similarly, in ethics, any assumptions about right or wrong (such as the alleged fact that torturing children is wrong) seems to depend upon further unwarranted assumptions about right or wrong (such as the alleged fact that the needless suffering of innocents is wrong).[55] Each of these three philosophical bootstrapping problems is characterized by a vicious and self-undermining regress.

Indeed, each of these three great bootstrapping problems suggests its own nightmare scenario involving a manipulator. When making claims about right and wrong, one assumes that one’s conscience has not been covertly manipulated to regard objectively wrong behavior as subjectively right. People assume that they live in a universe where torturing innocents is wrong and not just wrong because of how our brains are constituted. Similarly, when making claims about knowledge, one assumes that one is not a brain-in-a-vat. Just as one assumes that our conscience conforms to objective right and wrong, so too does one assume that our mental worldview conforms to the objective world. Finally, in ordinary responsibility practices, one assumes that a meta-controller has not exercised meta-control* over one. On the contrary, people naturally regard themselves as the authors of their own lives.

The Illusory Value of Free Will

The costs of hard compatibilism suggest that Watson’s limit divides two different kinds of free will. Hard compatibilist free will involves the extent to which an agent expresses capacities such as reasoning and intelligence. This free will also involves the extent to which this agent is free from coercion and constraint. Hard compatibilist free will also admits of degrees. One interesting consequence of this view is that beings might have greater freedom than humans do. For example, if free will is just a matter of one’s desires being integrated appropriately in a “mesh” then the desires of posthumans[56] might be integrated in more elaborate and sophisticated “meshes.”[57] According to Watson’s Limit, however, even the lives of posthumans may be the product of design.

The other kind of free will, being-causa-sui, is logically impossible and so its value is controversial.[58] Although it is logically impossible, one can understand how this kind of free will does not have the costs of hard compatibilism. This kind of free will promises, by stipulation, to guarantee that: one’s life is not absolutely predictable, that one’s life is not subject to meta-control, and that one is uniquely responsible for one’s actions.[59] Furthermore, this kind of free will fits the traditional understanding of free will as an all-or-nothing concept.

One can distinguish between these two kinds of free will on the following Diagram 2:

Diagram 2. Two kinds of free will

The placement of these two kinds of free will along the same spectrum can be misleading. To the extent that being-causa-sui is logically impossible, it cannot provide one more freedom than hard compatibilist free will. There is an important sense, however, in which these kinds of free will can be placed along the same spectrum.

Both kinds of free will can be placed along the same spectrum because they both promise something adaptive.[60] Consider two organisms, only one of which is guaranteed to be free from predictability, meta-control, and sharing responsibility. The organism which can secure these guarantees will be, other things being equal, more fit. Such an organism will be less vulnerable to predation and will have more control over its life.[61]

One method of rendering oneself less predictable and vulnerable to meta-control is to shed one’s self. For example, if an organism has a constitution which inclines that organism towards running in a predictable pattern, then one method by which that organism can protect itself is to shed that part of its constitution. To achieve perfect unpredictability and freedom from meta-control, however, one would need to abolish all of the self. In this way, the allure of free will is not unlike the health and beauty promised to young women by anorexia.[62]

Conclusion

In Freedom Evolves, Daniel Dennett writes that:

“And do you also advise diabetics to insist on ‘real’ insulin, instead of the ‘artificial’ stuff? If your real heart gives out some day, will you spurn an artificial substitute that can perform all the functions of your real heart? At what point does love of tradition turn into a foolish superstition? I claim that the varieties of free will I am defending are worth wanting precisely because they play all the valuable roles free will has been traditionally invoked to play. But I cannot deny that the tradition also assigns properties to free will that my varieties lack. So much the worse for tradition, say I.”[63]

We have seen that this is false: hard compatibilist conceptions of free will do not play “all the valuable roles free will has been traditionally invoked to play.” Hard compatibilism cannot guarantee: that agents are not absolutely predictable, that agents are not the victims of meta-control, and that agents are uniquely responsible for their decisions. Furthermore, one can argue that hard compatibilist views lack an error theory of incompatibilism and are less philosophically elegant than free will denial. Finally, hard compatibilism has an essentially revisionist character. By abandoning the “properties” that “tradition” assigns to free will, such as its being all-or-nothing, hard compatibilists defend the existence of free will in name only. The next section continues to explore the costs of hard compatibilism by critiquing the view of a prominent hard compatibilist—one who has given perhaps the most attention to the challenge of free will denial.

IV. Fischer

In three recent articles—“Responsibility and Manipulation,”[64] “The Transfer of Non-Responsibility,”[65] and “The Cards That Are Dealt You”[66]—John Fischer argues against free will denial. His arguments failed to persuade me but they deserve attention.[67] In particular, I will provide an error theory of hard compatibilism which attempts to explain why each of these arguments fail: hard compatibilists commit a sort of false self-ownership and fail to approach themselves with sufficient skepticism. In “Responsibility and Manipulation” Fischer argues against the incompatibilist conclusion of Derk Pereboom’s four-step argument. In “The Transfer of Non-Responsibility” Fischer argues against a crucial premise of the best arguments for incompatibilism (the Consequence Argument) and free will denial (the Basic Argument): the Transfer of Non-Responsibility (TNR). Finally, in “The Cards That Are Dealt You,” Fischer offers multiple arguments against free will denial and accuses of it of “a kind of metaphysical megalomania.”

Blameworthiness

The first criticism of Fischer’s view is not original—although it deserves more attention. Instead, I am including it here for the sake of completeness. This first criticism just notes a further cost of hard compatibilism: its inability to guarantee that an agent who is morally responsible for a given non-morally-neutral action is further praiseworthy or blameworthy for that action. Fischer resorts to this position to avoid the incompatibilist conclusion of Derk Pereboom’s four-step argument:

“In my view, further conditions need to be added to mere guidance control to get to blameworthiness; these conditions may have to do with the circumstances under which one’s values, beliefs, desires, and dispositions were created and sustained, one’s physical and economic status, and so forth. Professor Plum, it seems to me, is not blameworthy, even though he is morally responsible. That he is not blameworthy is a function of the circumstances of the creation of his values, character, desires, and so forth. But there is not reason to suppose that anything like such unusual circumstances obtain merely in virtue of the truth of causal determinism. Thus, I see no impediment to saying that Plum can be blameworthy for killing Mrs. White in Case 4. Note that there is no difference with respect to the minimal control conditions for moral responsibility is Cases 1 through 4—the threshold is achieved in all the cases. But there are (or may be, for all that has been said in Pereboom’s descriptions) wide disparities in the conditions for blameworthiness.”[68]

But, as Pereboom notes in his reply to critics, “it is not part of our intuitive picture that an agent can be morally responsible for an action, the action be morally wrong, and the agent knows it is morally wrong, while nevertheless the agent is not blameworthy.”[69] Indeed, if an agent can be morally responsible for an immoral action and yet not blameworthy for that action, the label “morally responsible” loses its significance. Again, a sufficiently revisionist compatibilism defends the existence of free will in name only. According to hard compatibilism, not only might free agents be perfectly predictable, the victims of meta-control, and share responsibility for their actions, but such free agents might not even deserve blame or praise for these actions.

The Transfer of Non-Responsibility

A more original reply to Fischer’s critique of free will denial might defend the Transfer of Non-Responsibility (TNR). According to TNR, if an agent is not responsible for both the event X, and the entailment X à Y, then that agent cannot be responsible for Y. The Transfer of Non-Responsibility is a crucial premise of the best arguments for both incompatibilism (the Consequence Argument) and free will denial (the Basic Argument).[70] For example, Fischer paraphrases this premise of the Basic Argument as follows: “If you do what you do because of the way you are, then in order to be ultimately morally responsible for what you do, you must be ultimately morally responsible for the way you are.”[71]

John Martin Fischer argues against TNR with examples such as the following:

[In “Erosion”], Betty plants ... explosives in the crevices of [a] glacier and [intuitively speaking, freely] detonates the charge at T1 causing an avalanche that crushes the enemy fortress at T3 [a result intended by Betty]. Unbeknownst to Betty... however, the glacier is gradually melting, shifting, and eroding. Had Betty not placed the dynamite in the crevices, some ice and rocks would have broken free at T2, starting a natural avalanche that would have crushed the enemy camp at T3.[72]

But such counter-examples beg the question against those who defend free will denial. Surely, such a person does not agree that Betty is “intuitively” morally responsible for the fact that the avalanche happens. So Betty’s not being morally responsible for neither the eroding glacier, nor for the fact that, if the glacier erodes then an avalanche will happen, does not violate TNR, because she is not responsible for the avalanche’s happening either.

Fischer’s example only seems to work because he is not using TNR to capture the intuitions which motivate it. In particular, he is using it selectively. The difference between apparently successful and unsuccessful examples depends upon the selection of X and Y, such that the agent is not responsible for X and X à Y, and therefore, given TNR, for Y as well. Fischer selects an eroding glacier for X and the following of an avalanche from this erosion for X à Y. Fischer does not consider other selections of X and Y which might undermine his position. But if one selects a different X one may ultimately vindicate TNR. In particular, one should focus upon features of Betty’s constitution such as her nature, character, or disposition. Her actions follow from this constitution. More importantly, Betty is not responsible for her original constitution (X) or the fact that her behavior follows from her constitution (X à Y). Once one recognizes that Betty is not responsible for this X, as opposed to Fischer’s X, the assumption that Betty is “intuitively” morally responsible for the avalanche is no longer uncontroversial.

This suggests that compatibilists are not doing justice to the sense in which agents cannot be responsible for their own constitutions. As Michael McKenna noted in his defense of TNR:

“Assuming determinism, the pertinent facts (consisting in the deterministic order of things) are not independent of an agent’s reasons for action, they constitute them! Therefore, at a deterministic world invoving a typical case regarding a judgment of moral responsibility, the case is relevantly like a one-path, not a two-path case.”[73]

Similarly, Galen Strawson, author of the Basic Argument, observes of compatibilists in general:

“Compatibilists claim that this is the right thing to say. They believe that to have free will, to be a free agent, to be free in choice and action, is simply to be free from constraints of certain sorts. Freedom is a matter of not being physically or psychologically forced or compelled to do what one does. Your character, personality, preferences, and general motivational set may be entirely determined by events for which you are in no way responsible (by your genetic inheritance, upbringing, subsequent experience, and so on). But you do not have to be in control of any of these things in order to have compatibilist freedom. They do not constrain or compel you, because compatibilist freedom is just a matter of being able to choose and act in the way one prefers or thinks best given how one is.”[74]

The Cards That Are Dealt You

Fischer confronts free will denial directly in “The Cards That Are Dealt You.” Unfortunately, the arguments he offers against free will denial all fail. Furthermore, they fail because Fischer’s view commits a sort of false self-ownership and fails to approach the self with sufficient skepticism. Consider the following analogies.

Poker Player

The first analogy that Fischer suggests provides his title. He writes “Our behavior may well be ‘in the cards’ in the sense that we simply have to play the cards that are dealt us.”[75] Here Fischer is comparing these “cards” with the initial “slant” or “bent” of an agent’s constitution (found in, for example, one’s genetics or childhood environment). There is a relevant difference, however, between playing the “cards that are dealt you” and taking responsibility for the initial “slant” or “bent” of one’s constitution. The difference is that, in the poker situation, the player’s constitution remains uninfected by the deal of the cards, whereas in the other situation, the “slant” or “bent” of one’s constitution does infect the agent’s constitution. In the poker situation, an agent can presuppose an authentic sense of self-ownership, and play the cards in a variety of ways. This is so because the self and the deal of the cards are kept separate. In the other situation, however, the “slant” or “bent” of one’s constitution determines how one will live one’s entire life. Indeed, Fischer compares an agent’s constitution to a situation (poker players plus the cards that are dealt them) which itself contains agents within it—betraying a hard compatibilist intuition that having free will is like having a homunculus inside of one’s mind. To appreciate just how disturbing this “slant” or “bent” ought to be, consider this revision of Fischer’s analogy: taking responsibility for one’s life, despite the “slant” or “bent” of one’s constitution, is like people playing poker where the deal of cards determines not just which cards each person has but also each person’s constitution and how they play their cards. We lose our intuition that these poker players can be responsible for how they play their cards. Their entire lives are dealt to them.

Astronaut

Fischer suggests a second analogy which also fails to appreciate the false sense of ownership:

“Further, just as an astronaut may still control the lift-off of the rocket, even though she did not build the platform that makes the launch possible (or ever have any control over the platform), we can be accountable for playing the cards that are dealt us, even if we did not manufacture the cards, write the rules of the game, and so forth.”[76]

Again, Fischer suggests an analogy for an agent’s constitution which itself contains agents within it. As agents existing within a larger agent, these suggest that Fischer’s view of having free will is like having a homunculus inside of one’s mind. Like the previous example, this one relies upon keeping this agent (“astronaut”) separate from the “slant” or “bent” of its constitution (“platform”). But the “slant” or “bent” of one’s constitution is not divorced or isolated from one’s self. Also like the previous example, this analogy can be revised to show just how disturbing the “slant” or “bent” of one’s constitution ought to be: taking responsibility for one’s life, despite the “slant” or “bent” of one’s constitution, is like an astronaut controlling the lift-off of a rocket, where the platform of the rocket determines not just the rocket’s starting position but also the astronaut’s constitution and how the astronaut will control the rocket.

The Inner Citadel

The rich irony of Fischer’s offering these examples is that he seems unaware of their violating principles which he swore to uphold earlier in his article. Indeed, he criticizes Saul Smilansky for suggesting that compatibilists seek an “Inner Citadel”:

“The mistake is to suppose that compatibilism seeks to identify an ‘Island of Control’—an Inner Citadel. It is better to think of compatibilism as concedeing from the beginning that we are thoroughly subject to factors entirely outside our control. Nevertheless, according to the compatibilist, we can still exhibit a meaningful and robust sort of control. It is not as if the compatibilist seeks to carve out a sphere of pure "internality" and immunity to arbitrariness, and then must be embarrassed to discover that the inner sanctum is not secure. He never thought that we needed such a place.”[77]

Yet the analogies that Fischer offers in support of hard compatibilism rely upon precisely this illusory picture of agency and autonomy! He places a poker player in “protective bubble” or “armored vehicles” uninfected by the deal of cards. Similarly, he circumscribes his astronaut by a fortress and isolates him in an “Inner Citadel” uninfected by the “platform” beyond his control. But this deal of “cards” and rocket “platform” cannot be relevantly similar to the “slant” or “bent” of an agent’s constitution because such a “slant” or “bent” does infect the agent. We do not deal our own cards.

Total Control

These analogies suggest a more rigorous argument against free will denial. According to Fischer those who deny the existence of free will are demanding a precise, and unreasonable, type of control:

“It is as if Strawson thinks of free and morally responsible agents as having ‘total control’. An agent has total control over X only if for any factor f which is a causal contributor to X and which is such that if f were not to occur, then X would not occur, the agent has control over f. But we have seen that total control is a fantasy. To have total control would be to have control over the sun's continuing to shine, the earth's not being hit by a meteorite, and so forth. The desire for total control is a reflection of a kind of metaphysical ‘over-reaching’, if anything is.”[78]

This definition may be too broad but not because the notion of control it describes is too demanding. According to free will denial people have a natural belief that they possess an amount of control over their lives; this intuition is mistaken just because the amount of control suggested is too demanding. There is one thing to say about the definition of total control. Total control is defined over a variable X. But those who deny the existence of free will do not do so because they lack total control over their bodily actions; rather, they do so because they lack control over their decisions.[79] For example, hurricane winds might blow persons against their will. Although these persons may be moved, their minds are untouched. There is more to say about total control in the context of Fischer’s argument about two axes.

Two Axes

Fischer offers a second, more rigorous argument against free will denial at the end of his article:

“Imagine, quite fancifully, that our agency is a connected set of dots—a horizontal line-segment from point b to point c. Now imagine a vertical line coming from below, with an arrow pointing toward the Agency Line. This line represents a causally necessary condition, such as the sun's shining; the sun's shining causally sustains and ‘sets the stage’ for the exercise of agency. Now add a line that is (like the Agency Line) horizontal, starting to the left of point b and with an arrow pointing toward b. This arrow represents a causally deterministic sequence issuing in b (the beginning of the exercise of agency). Suppose that the relevant agent is not in control of this antecedent causal sequence ‘pointing horizontally toward b’, just as he is not in control of the sun's shining. My question is this: if one is not troubled by the existence of the vertical line, why be troubled by the horizontal line? They are both the same in the sense that they represent ‘external’ factors that are entirely outside the relevant agent's control; in virtue of what is the horizontal line troubling in a way in which the vertical line isn't? A mere appeal to ‘externality’ will not distinguish the two lines—they are equally ‘external’ to the Agency Line. Similarly, the sun is ‘external’ to the agent in just the same way as the antecedent causal sequence—each equally impugns Total Control, and both introduces just the same sort of luck. Of course, this is not to say that there are no potentially relevant differences between causally sufficient and necessary conditions; but it is to issue a challenge to say what those differences consist in.”[80]

There are several answers to Fischer’s question “if one is not troubled by the existence of the vertical line, why be troubled by the horizontal line?” For one, to say that “one is not troubled by the existence of the vertical line” is not entirely true. There is something disturbing about the fact that people are dependent upon so many things, such as the sun, which are beyond their control. One can imagine how a person might desire to ensure that the sun always exists so that the person continues to exist as well.

More important, however, is the distinction between the horizontal and vertical lines. Fischer writes that “they are both the same in the sense that they represent ‘external’ factors that are entirely outside the relevant agent's control.” Indeed, both do represent things that are foreign to the self. Only one group of these foreign things, however, makes a contribution to the agent’s constitution. As Fischer notes, whereas the vertical line “sets the stage” for agency, the horizontal line determines how the agent acts upon this stage. For example, although the sun’s shining determines whether or not one exists, it is not clear to what extent it also determines what kind of agent one is.

Furthermore, those who deny the existence of free will do not demand total control over the sun per se. They might demand, what is only slightly less preposterous, just that, so long as the sun is outside of their control, it does not make a causal contribution to their selves. Rather than conquering the entire universe, people may wish to isolate themselves in their own Inner Citadels.

Even if this argument presses the distinction between the vertical and horizontal lines too far, however, an answer remains to Fischer’s question. Fischer accuses those who deny the existence of free will of “a kind of metaphysical megalomania.”[81] Perhaps, however, Fisher is not doing justice to widespread pre-philosophical intuitions about free will and moral responsibility. Fischer anticipates this objection:

“Of course, a full response to Strawson would need to explain how our intuitions about blame and fairness do not really lead us to such an excessively demanding and ‘exalted’ autonomy condition, and how our practices are justifiable in the face of the failure to fulfill this condition. I do not attempt such a response here. Rather, I am highlighting a dialectical move evidently not considered (or not given sufficient attention) by Strawson: that the sort of autonomy we might initially find attractive does not survive critical scrutiny—it is unduly demanding. In light of the implications of Strawson's picture of autonomy, one could go in either of two directions. One could say that such a picture is endorsed by commonsense but utterly impossible to fulfill. Or one could say that such a picture, being obviously and straightforwardly impossible to fulfill, cannot be the picture endorsed, upon reflection, by common sense. The latter possibility seems to be the approach suggested by Feinberg, and it seems to me to be the path recommended by a certain sort of philosophical maturity and wisdom. Be that as it may, my more minimal point (to which I would retreat if pressed) is simply that the latter approach is no less plausible than the former, given the considerations adduced by Strawson. His argument is, at best, incomplete at this critical juncture.”

But the term “free will” might be widely used without “reflection.” The answer to Fischer’s objection is not to deny that the consequences of this conception of free will are preposterous. Nor should one rely upon a relevant difference between the vertical and horizontal lines (although such a difference may exist). Rather the preposterous consequences of a given concept do not suffice to show that the concept is not widely used—but common usage is the ultimate arbiter of a term’s meaning. People may, in ordinary responsibility practices, widely use a demanding concept of free will because they fail to appreciate its preposterous consequences. Perhaps we are born metaphysical megalomaniacs.

Conclusion

John Fischer writes:

“The shift from a requirement of regulative control to a requirement of guidance control does not in itself provide a decisive defense of compatibilism about causal determinism and moral responsibility. But it does shift the debate into a dialectical terrain considerably more hospitable to compatibilism. If we do not need genuine metaphysical access to non-actual possibilities, a compatibilist need not deny the fixity of the natural laws. The semicompatibilist conceives of the role of hypothetical scenarios in a distinctive way; thus, the theoretical role of these scenarios is fundamentally different in traditional compatibilism and semicompatibilism. To mix elements of the older ways of thinking with the new only breeds confusion.

I proclaim, then, with all due modesty: Semicompatibilism is the Free Will Revolution. Agency theorists of the world, unite!”[82]

But semi-compatibilism, which asserts the compatibility of determinism and moral responsibility while denying the Principle of Alternative Possibilities, is not very revolutionary. Semi-compatibilism is a movement within compatibilism and, like all varieties of compatibilism, defends a largely orthodox and uncontroversial view of agency. In particular, semi-compatibilism soothes the philosophical doubts people have about free will and moral responsibility. “Who’s afraid of determinism?” the semi-compatibilist might ask. Other hard compatibilists, such as Daniel Dennett and Gary Watson, defend arguably more radical views.

This section has shown that there is a significant cost to defending orthodoxy. In particular, hard compatibilists cannot guarantee that agents who are morally responsible for a non-morally-neutral action are also praiseworthy or blameworthy for that action. Semi-compatibilists also fail to appreciate the Transfer of Non-Responsibility and their arguments against that principle do not succeed. Finally, the analogies and arguments that Fischer employs against free will denial in “The Cards That Are Dealt You” rely upon a relevant similarity between relevantly dissimilar things—and betray the semi-compatibilist’s ironic conception of agency as involving a homunculus or “Inner Citadel.” Hard compatibilists makes these mistakes because they fail to approach the self with sufficient skepticism.

The following section focuses upon a concept which Fischer has considered but perhaps given too little attention. Those who deny the existence of free will may be metaphysical megalomaniacs because they make too stringent demands for control; they may also be metaphysical megalomaniacs because they make too stringent demands for what kind of agent to be: perfect agents. The notion of a perfect agent suggests a genuine Free Will Revolution. If hard compatibilists can ask “who’s afraid of determinism?” then those who deny the existence of free will can ask “who’s afraid of creeping exculpation?”

V. Perfect Agents

“Now Free Will is obviously not included in our common ideal of physical and intellectual perfection: and it seems to me also not to be included in the common notions of the excellences of character which we call virtues: the manifestations of courage, temperance, and justice do not become less admirable because we can trace their antecedents in a happy balance of inherited dispositions developed by a careful education.”

—Henry Sidgwick[83]

Reflecting upon the causes and cures of moral imperfection tests one’s belief in the existence of free will and moral responsibility. Tradition regards the causes of moral imperfection as the agents themselves; tradition regards the solution to moral imperfection as punishment. Both of these orthodox answers have severe limitations. The causes of moral imperfection run through agents and into their pasts. Answering that agents misbehave because of who they are does not answer the question of how agents came to be who they are. Similarly, punishment can improve behavior to only a modest extent and inflicts a pain cost upon the punished.

This section will explore how the concept of a perfect agent tests one’s belief in free will and moral responsibility. The section begins by discussing three different kinds of perfect agents: the confident agent, the strongly reasons-responsive agent, and the natural Epictetan. Consideration of these three kinds of perfect agents will suggest an explanation for why their slavery is especially conspicuous.[84] The section continues by distinguishing between two kinds of perfection. A failure to distinguish between these two kinds of perfection may cause philosophers to reach mistaken answers to a variety of problems. These problems include whether aging is a disease and the problem of evil.

Three Kinds of Perfect Agents

This article will refer to agents who are never uncertain about what to do, especially those who are never uncertain about what is right, as perfect agents. Humans are imperfect agents. In this sense, perfect agents, like many meta-controllers (or amazingly skilled “neurosurgeons”), are non-human. Also like the concept of a meta-controller, the concept of a perfect agent, in the context of the free will problem, may be illuminating. This section will consider three kinds of perfect agents: the confident agent, the strongly reasons-responsive agent, and the natural Epictetan. The creator of the confident agent argues that such an agent is intuitively free. The creator of the reasons-responsive agent would claim that such an agent is free but does not subject this claim to much examination. Finally, the creator of the natural Epictetans argues that such agents are intuitively not free. These different conclusions may be due to differences in the three kinds of perfect agents. This section will: argue that the former two creators are mistaken, argue that the latter creator is correct but for a mistaken reason, and suggest a different explanation for why the slavery of perfect agents is especially conspicuous.

The Confident Agent

In a recent article, “Close Calls and the Confident Agent: Free Will, Deliberation, and Alternative Possibility,” Eddy Nahmias considers one sort of perfect agent: the confident agent. Nahmias argues that the free will problem involves two conflicting intuitions:

“One intuition locates free will in our ability to deliberate effectively and control our actions accordingly: the ‘Deliberation and Control’ (DC) condition. The other intuition is that free will requires the existence of alternative possibilities for choice: the AP condition.”[85]

The article continues by considering a confident agent who deliberates about choices but always feels confident about the outcome of this deliberation. Such a confident agent maximizes the DC condition but never experiences “close calls” and so questions the importance of the AP condition. Nahmias relates his argument as follows:

1) Some incompatibilists (e.g., van Inwagen, O’Conner, Kane) suggest that free will requires alternative possibilities (AP) and that the AP condition requires the existence of close calls.

2) The confident agent does not face close calls.

3) So, according to them, the confident agent could not have free will.

4) But, intuitively, the confident agent does have free will.

5) So, contra these incompatibilists, it seems free will does not require alternative possibilities.[86]

Nahmias recognizes, however, that step four is controversial and so he supports his claim by answering six objections.[87] Objection five is the one which concerns this article:

“Objection 5: One might also object that the confident agent, as I describe her, may have free will but only if she does not exist within a deterministic universe or only if she has agent causal powers. This objection suggests that I have begged all the important questions about the traditional problem of free will.”[88]

In answering this objection, Nahmias is quick to emphasize that he is not arguing for compatibilism but rather is suggesting an explanation for why incompatibilists seek alternative possibilities:

“However, some incompatibilists argue that the problem with determinism is not that it threatens the AP condition but that it conflicts with an agent’s ability to be the ‘ultimate source’ of her actions. So, while the confident agent thought experiment may suggest that the AP condition is not essential for free will, it does not challenge this ‘source incompatibilism’ since there is no reason to think she is the ultimate source of her actions (at least if she exists in a deterministic universe). If my thought experiment helps motivate more incompatibilists to take this position, as have Frankfurt cases, then that will be a significant result, since the arguments for source incompatibilism are, I think, even more controversial than those relying on the AP condition. I also think the confident agent can help us see why this is. Even though she is not described as initiating some choice such that there are no complete causal explanations for her choice, she clearly endorses the reasons that are essential proximate causes of her actions. As long as those reasons themselves arise through her own deliberative process and are not manipulated in some autonomy-undermining way, it is not clear why we should demand that, if she is to be considered free, those reasons themselves cannot have sufficient prior causes.”[89]

Indeed, Nahmias’ thought experiment does not challenge this “source incompatibilism” and so the truth of his fourth premise (“But, intuitively, the confident agent does have free will.”) remains controversial. This article has suggested its own error theory of libertarianism which grants precisely this “significant result”: the demand for alternative possibilities represents a misguided attempt to be causa-sui. From this perspective, one might regard perfect agents, not as especially free, but as most clearly enslaved.

Strong Reason Responsiveness

Like Nahmias, John Fischer considers the concept of a perfect agent and agrees that such an agent would have free will. This perfect agent is “strongly reasons-responsive” where strong reasons-responsiveness means:

“(SRR): Suppose that a certain kind K of mechanism actually issues in an action. Strong reasons-responsiveness obtains under the following conditions: if K were to operate and there were sufficient reason to do otherwise, the agent would recognize the sufficient reason to do otherwise and thus choose to do otherwise and do otherwise.”[90]

So this perfect agent is one who never fails to respond to good reasons. Fischer and his coauthor consider strong reasons-responsiveness, however, largely in order to argue against its being a necessary condition for moral responsibility. Instead, they argue that a less demanding amount of responsiveness, moderate reasons-responsiveness, is necessary. Strong reasons-responsiveness, on Fischer’s view, is a sufficient but not necessary degree of responsiveness for moral responsibility. Fischer and his coauthor do suggest, however, that strong reasons-responsiveness may be necessary for something else:

“Strong reasons-responsiveness may be both necessary and sufficient for a certain kind of praiseworthiness—it is a great virtue to connect one’s actions with the contours of value in a strongly reasons-responsive way. But, of course, not all agents who are morally responsible are morally commendable (or even maximally prudent).”[91]

So strong reasons-responsiveness, on Fischer’s view, is not necessary for moral responsibility. Imperfect agents might be nevertheless be morally responsibility. This conclusion, which Nahmias and Fischer reach, is remarkable for at least three reasons. For one, imperfect agents, on Fischer’s view, might be blameworthy for immoral actions despite the fact that they are not morally responsible for being imperfect agents (X) and if they had done otherwise then they would have been a more perfect agent (~Y à ~X). There is a tension here between Fischer’s view and the Transfer of Non-Responsibility principle—which this article has shown survives Fischer’s critique. Secondly, Fischer does not suggest that moderately reasons-responsive agents ought to become strongly reasons-responsive. Instead, his view gives one the impression of moderately reasons-responsive agents praising and blaming each other without ever becoming more than moderately reasons-responsive. They do not aspire for perfection. Finally, Fischer’s discussion of strongly reasons-responsive agents is remarkable because he reaches the opposite conclusion that Strawson does about similar agents: natural Epictetans.

Natural Epictetans

John Fischer writes that “Despite my disagreement with Strawson about ultimacy, I have considerable admiration for his work on free will and moral responsibility.”[92] Indeed, Fischer agrees with Strawson about a controversial position within the free will debate: subjectivism. According to this subjectivism, to actually possess free will, agents must believe that they possess it. This is where the similarities end. Although Fischer and Strawson have both given attention to perfect agents and free will denial, they have reaches opposite conclusions. Strawson concludes that free will does not exist and that a kind of perfect agent would lack it; Fischer concludes that free will does exist and that a kind of perfect agent would have it. In this respect, Strawson’s view is the more revolutionary.

After relating the succinct and devastating Basic Argument in Freedom and Belief, Galen Strawson explores other issues involving agency. In particular, Strawson wishes to defend a claim with which Fischer agrees: subjectivism. To defend this claim Strawson suggests the concept of a “natural Epictetan” which is similar to the strongly reasons-responsive agent:

“Imagine an enormously congenial world inhabited by a race of gifted, active creatures. They fulfil all the agent-structural conditions of U-freedom so far proposed, and the (attitudinal) engagement condition. What is unusual about them is that they are never undecided in any way. They are fully—richly—able to choose, but they never hesitate at all about what to do. They never ponder alternatives, although they are perfectly well aware of them. They never consciously deliberate about which ends to pursue or about how to pursue them (they have no need to) and always succeed in doing what they want to do.

These are the Natural Epictetans—never failing, never disappointed in their congenial world, always able to do what they want to do because always wanting to do only what they are able to do.”[93]

According to Strawson, however, such perfect agents obviously lack free will and moral responsibility. They are like “like creatures that have ears but live in a soundless world.” Indeed, Strawson attributes the slavery of the natural Epictetans to their lack of belief in their own freedom:

“The proposal, then, is that the Natural Epictetans aren’t U-free agents although they are fully self-conscious, engaged agents: they can’t be U-free agents—even if they fulfil the causa sui condition—because they have no sense of what U-freedom is. They fail to fulfil condition [10], the belief in freedom condition; they fail to figure themselves as U-free.”[94]

Strawson even suggests that if an agent could be causa-sui, in some paradoxical sense, without believing that it was free, then this lack of belief would suffice to destroy its freedom. But belief in one’s freedom does not seem to be a requirement of one’s being free. Strawson notes that Thomas Nagel raised this objection:

“This is a difficult claim to assess, since it involves a counter-possible conditional: but it doesn’t seem right to me. If true responsibility [U-freedom] were possible, couldn’t someone be deceived (even self-deceived) about whether he had it? Couldn’t he act with the illusion that all this was just happening to him, while actually it was his doing, and he was fully responsible for [U-free in] a choice which saved or failed to save ten other people from torture?”[95]

Furthermore, it is not clear why the natural Epictetans cannot believe that they are free. Perhaps Strawson is suggesting that a belief in one’s freedom implies indecision. Although Strawson feels confident that this lack of belief explains the natural Epictetans’ conspicuous slavery, he later entertains the Epictetans’ lack of indecision as the explanation: “If one asks about the deep sources of belief in U-freedom, experience of indecision seems far more important than experience of constraint or hindrance.” Here Strawson agrees with Nahmias about the importance of indecision but reaches the opposite conclusion about agents who never experience it.

Conspicuous Slavery

The explanation for the different conclusions which Strawson and Nahmias reach may be that confident agents and the natural Epictetans are not identical.[96] The more likely explanation, however, is that Nahmias does not share Strawson’s agreement with the importance of sourcehood. Indeed, Strawson boasts that his Basic Argument does not rely upon the incompatibilist intuitions which Nahmias’ confident agent was intended to undermine. Nahmias may also not share Strawson’s subjectivism. Fischer, however, does share Strawson’s subjectivism. Considering this mutual subjectivism, I suspect he may sympathize with Strawson’s claim that the strongly reasons-responsive agent, or the natural Epictetan, is not free. In summary, Fischer and Nahmias fail to explain why Strawson regards perfect agents as so clearly enslaved. This article will here suggest its own explanation—which differs from Strawson’s.

There is an essential similarity between these perfect agents which does undermine their freedom and responsibility. Confident agents, strongly reasons-responsive agents and the natural Epictetans are all kinds of perfect agents: they always know exactly what they want. The slavery of such agents is especially conspicuous, not because they do not believe in their own freedom, but because this certainty highlights their lack of participation in the creation of their own constitutions. Uncertainty, like alternative possibilities, disguises the fact that an agent is not causa-sui. An observer might think, according to pre-philosophical intuitions, that uncertain agents are in the process of creating themselves ex-nihilo. In contrast, that a perfect agent receives, rather than chooses, its constitution is transparent. Perfect agents project and broadcast the wishes, goals, and preferences which are given to them. Furthermore, they show no sign of doubting, struggling with, or wishing to revise, these constitutions. The uncertainty of imperfect agents, according to pre-philosophical intuitions, disguises their slavery. But the slavery of perfect agents is plain.

Two Kinds of Perfection

These considerations of perfect agents suggest an explanation for a remarkable phenomenon: the unwillingness to criticize obvious, but intractable, design flaws. To be more precise, suppose that an observer has a set of goals. Suppose further that this observer is evaluating a given work. The phenomenon of interest here is observer’s tendency, if the work has intractable design flaws, to evaluate the work according to how well it conforms to its specification and not according to how well it conforms to the observer’s own goals. The intractability of design flaws inclines one to turn a blind eye towards them. This section will explore the unwillingness to criticize obvious, but intractable, design flaws by considering two examples: aging and the problem of evil.

Aging

First, this phenomenon explains the resistance to categorizing aging as a disease. Stating necessary and sufficient conditions for a disease is a contentious project. This project seems to have taken place most vigorously during the seventies.[97] But progress in biogerontology has renewed interest in the question of whether aging is a disease.[98] Nevertheless, aging ought to be considered a disease according to largely uncontroversial criteria.

Consider the following simple definition of disease offered by Lester King in “What Is Disease?”:

“Disease is the aggregate of those conditions which, judged by the prevailing culture, are deemed painful, or disabling, and which, at the same time, deviate from either the statistical norm or from some idealized status.”[99]

According to these criteria, aging is a disease. The relevant clause here refers to “some idealized status.” King includes this clause because he entertains the possibility of universal diseases.

Similarly, consider the following criteria offered by Caroline Whitbeck, in “Four Basic Concepts In Medical Science.” She writes that “A disease is any type of psycho-physiological process such that:”

1.1 People wish to be able to prevent or terminate the process because it interferes with the bearer’s psycho-physiological capacity to do those things that people commonly wish and expect to be able to do;

1.2 Either the process is statistically abnormal in those at risk or people have some other basis for a reasonable hope of finding means to prevent or effectually treat the process; and

1.3 The process is not also necessary for doing anything that people commonly want and expect to be able to do.[100]

Aging is a disease according to these criteria. Like King, Whitbeck entertains the possibility of universal diseases that are not statistically abnormal. She hinges this possibility, however, upon the reasonable hope that one may eventually treat or prevent such a universal disease. So although, at the time she wrote her article, Whitbeck may not have considered aging itself a disease, recent work has made this classification more controversial.[101] Whitbeck even addresses aging explicitly:

“In contrast, when we consider human diseases, we count as diseases some processes which are statistically normal in those at risk. Furthermore, medical science has come to classify as diseases, processes which were previously considered normal. Consider especially those universal or near universal changes which in former times were considered to be normal consequences of aging, and which now are regarded as diseases in spite of their universal presence in older people. In some instances the processes were seen to bear some resemblance to others which we already termed ‘diseases’ when they occurred in younger people, in other instances, such as arteriosclerosis, increase in severity led epidemiologists to focus on the causes of this change and hence to raise the possibility of delaying or retarding the progress. In all cases we came to think of the process as one which we might be able to prevent or treat, and therefore no longer a necessary feature of the human life cycle. In terms of the foregoing definition of disease we came to regard the conditions outlined in criterion 1.2. as having been met.”

In “Health as a Theoretical Concept” Christopher Boorse offers the following criteria for a disease:

1. The reference class is a natural group of organisms of uniform functional design; specifically, an age group of a sex of a species.

2. The normal function of a part or process within members of the reference class is a statistically typical contribution by it to their individual health survival and reproduction.

3. Health in a member of the reference class is normal functional ability: the readiness of each internal part to perform all its normal functions on typical occasions with at least typical efficiency.

4. A disease is a type of internal state which impairs health, i.e. reduces one or more functional abilities below typical efficiency.

According to these criteria, aging is not a disease. But Boorse considers this a “failure” or “anomaly” and attempts to revise his account accordingly. Like Whitbeck, he addresses aging explicitly:

“The revised definition covers conditions like lung irritation and provides an alternate explanation of tooth decay. What there cannot be, on this view, is a universal genetic disease.

Exactly that, however, may be implied whenever medical authors list progressive dysfunctions of normal aging as diseases. When senile decline of function is caused from within, our account will not allow it to be a disease. That is because of the age-relativity which we built into the account to reflect differences between child and adult. Apart from childhood, one might be tempted to take the adult as the species type and old age as its disintegration. Yet the same functional limitations viewed as diseases in old age may count as normal in childhood. Much of senility is only regression to earlier stages of development. The puzzle is why old age is not always seen as a stage with its own statistical norms of healthy functioning. Lacking a solution to this puzzle, our account ends up differing from some medical sources over whether minor deformities and normal aging constitute a disease.”

Before offering these flawed criteria, Boorse considers several themes associated with diseases: value; treatment by physicians; statistical abnormality; pain, suffering, and discomfort; disability; adaptation; and homeostasis. Aging is disvalued, causes pain, suffering, and discomfort, and causes disability. Human aging may soon be treatable just as aging is treatable in several other species.[102] Deviation from homeostatis is an unpopular criterion for disease. Defining aging has been controversial because of the remaining themes: statistical abnormality and adaptation. Boorse recognizes both of these themes to be problematic and abandons them.

The resistance to classifying aging as a disease represents an unwillingness to criticize an obvious, but intractable, design flaw. Humans might have experienced negligible senescence. God or natural selection specified, however, that humans are supposed to age. An observer might suggest that an aging human better achieves their goals than a potentially immortal one. But this claim is doubtful. Aging causes so much agony, and so little good, one might suspect instead that the potentially immortal person would better achieve. Those who appose life extension seem to compromise their own goals because they are unwilling to criticize the specification of human nature. Furthermore, this unwillingness is convenient because criticism would be futile. For millions of years, the problem of aging has been intractable. Just as the opponents of life extension turn a blind eye to the obvious, but intractable, design flaw of aging so too are they unwilling to admit another flaw: evil.

The Problem of Evil

Derk Pereboom notes: “The free will theodicy in systematized form dates back at least to St. Augustine, and remains the most prominent of all theodicies.”[103] So the problem of evil is intimately related to the problem of free will. Indeed, meta-controllers demonstrate the costs of hard compatibilism and God is the quintessential meta-controller. The relationship between the problem of evil and the problem of free will becomes more apparent when one considers the unwillingness of many to criticize obvious but intractable design flaws.

There is a curious tension between the imperfect nature of humans and traditional religious beliefs. One might wonder why God—if he exists—did not create perfect agents. Human imperfection is intimately related to the problem of evil because the cause of much of the evil in the world is human. Indeed, one might wonder why God did not create humans to be more like strongly reasons-responsive agents. Defenders of religious beliefs have suggested a number of answers, of varying plausibility, to the problem of evil. For example, Derk Pereboom addresses Alvin Plantinga’s free will theodicy in his article “The Problem of Evil”:

“Plantinga argues that it is possible that this claim is false. For in his view it is possible that (God knows that) every possible person—i.e. every person-essence— has transworld depravity. For such an essence to suffer from transworld depravity is for it to be such that if God had created the person, and had given her significant freedom, then no matter what circumstances God were to place her in, she would go wrong with respect to at least one action, so long as God left her significantly free. Consequently, if an essence suffers from transworld depravity, it is not within God’s power to weakly actualize a possible world in which the corresponding person is significantly free and yet never makes a wrong free decision. But if it is possible that every relevant essence suffers from transworld depravity, then no matter what world featuring significantly free beings God weakly actualizes, there will be evil in that world.”[104]

The free will theodicy relies, however, upon a libertarian conception of free will.[105] This must be the case because God could obviously have perfected hard compatibilist agents. For example, he could have made them like the strongly reasons-responsive agent. Indeed, God could have made hard compatibilist agents godlike.[106] But, as Derk Pereboom notes, the existence of libertarian freedom is most controversial.[107] Furthermore, even if humans had libertarian freedom, it is not clear that this freedom could enhance or enable their moral responsibility.[108]

There are also non-traditional responses to the problem of evil. For example, the defenders of traditional religious beliefs might concede that God is not omnipotent to allow that God and evil can coexist. Again, Derk Pereboom notes the problem that the problem of evil remains because humans demonstrate an ability to improve themselves:

“From the point of view of traditional theism, such a position faces several problems. One is that if God’s lack of power alone (and not in addition some countervailing evil force) explains why he did not in the past prevent diseases such as smallpox, then since we can prevent smallpox now, we are in some respects now more powerful than God, at least than he was in the past. And since we are not worthy of worship, God’s worthiness to be worshiped is thus rendered doubtful. Another problem is that if God’s lack of power explains why he did not prevent smallpox, or the people in the Lisbon earthquake of 1754 from being crushed by the rubble of the churches they were attending that Easter Sunday morning, then how could he be powerful enough to create bacteria and viruses or wood and stones, let alone the entire universe? Furthermore, if God is not powerful enough to be the creator, the reasons for believing in God expressed by the teleological argument will have to be relinquished.”[109]

By curing smallpox, the human species improved itself and therefore seems more powerful than this non-omnipotent God. A third response to the problem of evil is skeptical theism. Skeptical theism just doubts whether evil is evidence that God exists. Derk Pereboom addresses the skeptical theism of Peter Van Inwagen—a prominent libertarian. Again, the problem of evil remains because humans demonstrate an ability to improve themselves:

“But are we really in no position to assess the probability of this defense on the existence of God? A misgiving about this claim arises from the fact that through fairly recent advances in technology and medicine we have prevented a significant amount of suffering, and, obviously, we have by these means prevented this suffering without introducing massive irregularity. But if we are now able in this way to prevent suffering, it would seem that God could have done so long ago without introducing massive irregularity.”[110]

Each of these responses shows that humans might have been perfect agents. God specified, however, that humans were supposed to be imperfect. An observer might suggest that a libertarian agent who commits evil better achieves their goals than a morally perfect hard compatibilist agent. But again one might doubt this claim. Since the helpfulness of indeterminism for moral responsibility is so controversial, and since the evil that morally perfect agents might avoid or prevent is so great, one might suspect instead that the morally perfect hard compatibilist agent would better achieve one’s goals—and so God does not exist. As in the previous example, this unwillingness to criticism an obvious design flaw is convenient, because criticism would be futile. Criticizing human nature would improve this nature nor convince God to intervene.

Conclusion

This section has introduced the concept of a perfect agent. First, the section explored three kinds of perfect agents and suggested that their perfection highlighted their slavery. Second, the article noted a curious phenomenon: the unwilling of many to criticize obvious, but intractable, design flaws. This phenomenon has led people to reach mistaken conclusions about several problems, including whether aging is a disease and whether people may be morally responsible for evil. The concept of a perfect agent is also important, however, because it suggests an alternative to the tradition of blaming and punishing agents for their actions: using meta-control to make perfect agents whose slavery is transparent. Indeed, this phenomenon helps explain the popularity of belief in free will: this belief will remain intractable so long as society cannot rehabilitate criminals without punishing them.

VI. The Benefits of Free Will Denial

“If you do not find a remedy to these evils, it is a vain thing to boast of your severity in punishing theft, which though it may have the appearance of justice, yet in itself is neither just nor convenient. For if you suffer your people to be ill-educated, and their manners to be corrupted from their infancy, and then punish them for those crimes to which their first education disposed them, what else is to be concluded from this, but that you first make thieves and then punish them?”

—Thomas More[111]

“If such is done of ‘disease,’ why not of ‘crime’? Not only is it clear that crime is a disease whose root is in heredity and environment, but it is clear that with most men, at least when young, by improving environment or adding to knowledge and experience, it is curable. Still with the unfortunate accused of crimes or misdemeanors, from the moment the attention of the officers is drawn to him until his final destruction, everything is done to prevent his recovery and to aggravate and make fatal his disease.”

—Clarence Darrow[112]

This section sketches what will replace our belief in free will and argues that this replacement has distinct benefits.[113] The section begins by arguing that immoral actions are symptoms of disease. The discussion continues by noting how meta-controllers suggest a cure to this disease. Next, the section anticipates and responds to several objections: that this conception of disease is too broad; that this cure has no general deterrence function; that this cure violates the dignity, autonomy, and personal identity of the controllee; that this cure is impossible and that society will reject it; and that this cure does not satisfy the demands of retributivism. Finally, the article will explore the intimate relationship between transhumanism, consequentialism, and free will denial, and argue that there is a natural harmony between them.

Crime is a Symptom of Disease

So begins Antony Flew’s classic “Crime or Disease?” in which he wrestles with this paradox and rejects its conclusion.[114] This article will argue that Flew was wrong: the brain conditions which produce immoral behavior, like aging, represent a disease. The article will refer to these conditions as immoral behavior* for brevity while noting that immoral behavior is just a symptom of the underlying disease. Consider, again, Lester King’s definition of disease:

Immoral behavior* deviates from the statistical norm and certainly from “some idealized status.” The weakest argument for immoral behavior* as a disease is that such a condition is painful or disabling. Indeed, immoral behavior* is typically painful or disabling for another and this seem to distinguish it from diseases. In response, one might argue that the conception of disease should include those conditions which are painful or disabling to others. Alternatively, one might argue that immoral behavior* is disabling: it disables the person’s capacity to be moral. The requirement for painfulness or disability, however, may not be necessary. Consider, again, Caroline Whitbeck’s criteria for a disease:

1.1 People wish to be able to prevent or terminate the process because it interferes with the bearer’s psycho-physiological capacity to do those things that people commonly wish and expect to be able to do;

1.2 Either the process is statistically abnormal in those at risk or people have some other basis for a reasonable hope of finding means to prevent or effectually treat the process; and

1.3 The process is not also necessary for doing anything that people commonly want and expect to be able to do.[116]

Unlike King’s definition, Whitbeck’s does not require painfulness or disability. Instead, she only requires that people “wish to be able to prevent or terminate the process because it interferes with the bearer’s psycho-physiological capacity to do those things that people commonly wish and expect to be able to do.” Most, but perhaps not all, people would wish to rid themselves of a condition which made them susceptible to behavior they considered immoral. Note, however, that Whitbeck does not require that the bearers themselves wish to prevent or terminate the process. This requirement is not necessary for a disease: cancer patients who love their tumors still have a disease (perhaps more than one). Furthermore, immoral behavior* does tend to interfere with the bearer’s psycho-physiological capacity to do those things that people commonly wish and expect to be able to do: living in peace and harmony with society without provoking its retaliation. Finally, immoral behavior* is a process which people have some a “reasonable hope of finding means to prevent or effectually treat.” This section will later suggest this cure: meta-control.

Flew

Antony Flew suggests his own criteria for disease but does not consider crime (or immoral behavior* more broadly) to satisfy these criteria. Crime must be “such as to be regarded by the patient as in itself and by his standards bad” and “be something which of its very nature tends to inhibit capacities”[117] Again, most people would consider immoral behavior* to be “in itself” and by their standards “bad.” Even if some people do not consider immoral behavior* to be “in itself” and by their standards “bad,” however, this requirement is not necessary for a disease. Furthermore, immoral behavior* arguably is something which “of its very nature tends to inhibit capacities.” Immoral behavior* inhibits one’s moral capacity.

Flew critically examines five reasons for why a person might consider crime to be the symptom of disease:

1. “there are obvious and uncontroversial analogies between many moral and quasi-moral faults… and common typical physical diseases,”

2. “someone might adopt this paradox… in a campaign for a clinical scientific approach to the study of crime and its prevention,”

3. “the adoption of the same paradox as a slogan by those who want to cut out indignation and moralizings and to treat criminals as individual, human problems,”

4. “Someone may say or suggest that ‘all crime is a symptom of mental disease,’ because he believes that ‘free will is an illusion,’ that no one ever can help doing what he does, and so on.”

5. “As a psychiatrist I see all crime as a symptom of disease.”

The fourth reason is the one that concerns this article. In arguing against the validity of this reason for saying “crime is a symptom of disease,” Flew makes several points. This section will examine three of these points: determinism implies predictability but not fatalism, determinism loses its string if one adopts a conditional analysis of ability, and the Argument of the Paradigm Case shows that free will must exist.

Fatalism

Flew concedes that hard compatibilism cannot guarantee that agents are not perfectly predictable but he is quick to emphasize that predictability does not imply fatalism:

“But determinism only commits us to saying that it was in principle possible to predict whether or not he would try and what he would succeed in doing. It does not commit us to—thought it is compatible with—fatalism: saying that whatever he did or tried to do, the upshot would have been the same. Thus, even complete determinism is quite compatible—as complete fatalism manifestly is not—with the fact that certain things are or were alterable by human effort, and particularly that there are some cases whether people could help doing what they did, or are going to do.”[118]

There are two points to make about Flew’s argument here. For one, Flew is willing to grant that hard compatibilism is compatible with absolute predictability but he fails to recognize the more severe costs of hard compatibilism: that a meta-controller may have designed an agent’s entire life and therefore share in responsibility for it.[119] Secondly, a little philosophical rigor shows that Flew’s argument against fatalism is not quite persuasive. Flew has shown that human effort can avoid the consequences of events in non-actual worlds. But this is not the fatalist’s claim. According to fatalism, “we are powerless to do anything other that what we actually do.”[120] The fatalist argues that human effort cannot avoid only those events in the actual world. Flew gives no reason for suggesting that human effort can avoid such events. This is the fatalist’s claim and so fatalism may survive Flew’s challenge. As a hard compatibilist, Flew seems to make this mistake because he refers to non-actual worlds in evaluating an agent’s capabilities.

The Conditional Analysis

Finally, Flew defends the conditional analysis of ability.

“Thus to show that John Doe “could not help himself,” that “he could not have done otherwise,” it is by no means sufficient to show that his behavior (presumably like that of everyone else, always) was on this occasion in principle predictable. It is not even relevant: for the sense in which the determinist says (none of us) “could have done otherwise” is not at all the everyday sense of “could have done otherwise.” The former is a question of predictability. The latter of would he have been able to do otherwise if he had tried: or was it beyond the reach of a man of his height; the capacity of one of his I.Q., etc.”[121] (italics added)

The conditional analysis of ability is “controversial,” however, and “perhaps untenable.”[122] In particular, it may lead to an infinite regress which the hard compatibilist cannot satisfy.[123] Considering this controversy, the conditional analysis is shaky ground on which to rest the conclusion that crime is not a symptom of disease.

The Argument of the Paradigm Case

One argument Flew makes against saying that “crime is a symptom of disease” because “free will is an illusion” is the Argument of the Paradigm Case:

“Here I appeal to what has usefully at last been named, the Argument of the Paradigm Case. As the meaning of expressions such as “of his own free will” is and must ultimately be given by indicating cases of the sort to which it is pre-eminently and by ostensive definition applicable, and not in terms of some description (which might conceivably be found as a matter of fact not to apply to anything which ever occurs); it is out of the question that anyone could now discover that there are not and never have been any cases to which these expressions may correctly be applied. Though of course he may well discover—what is certainly the case—that the expression has sometimes been misconstrued or mistaken to imply what it does not imply.”[124]

Free will is a philosophical term of art, however, and its meanings may vary with its usage. The phrase “of one’s own free will” is a popular phrase that implies little more than legal responsibility—the absence of legal defenses such as self-defense or uncontroversial mental illness. Such a usage of the term “free will” does not settle the question of whether free will or moral responsibility exist so much as it sidesteps this question and assumes that they do. The term “free will” is often used to refer to something the existence of which is more controversial: the freedom relevant condition for moral responsibility in a philosophical, as opposed to legal, sense. People use the phrase “of one’s own free will” without considering the challenge of determinism; philosophers use the term “free will,” however, in precisely this context. Furthermore, Flew seems to be conflating these two meanings because he later addresses the issue of determinism. For these reasons, the Argument of the Paradigm Case has not persuaded many philosophers that free will exists. Consider the conclusion to “The Paradigm Case Argument and the Free-Will Problem” by Arthur C. Danto:

“So the quarrel between the free-willist and his metaphysical enemy in no way resembles an argument between two plain men who might debate over whether or not Smith married of his own free-will. One need only imagine the sorts of reasons they might give, the facts that they might cite. Then one should read through the controversy between Hobbes and Bishop Bramhall. There is, so far as I can see, not a single argument, not a single reason there which would be of use to those plain men in settling their dispute. To introduce any of the Hobbes-Bramhall sorts of arguments would warp the fabric of the plain men’s controversy; it would bring it to a half. The disputants could then go on to argue metaphysics, or they could return to the history of Smith’s matrimony. But not both together.

One might now ask: if we can truly decide, at least in principle, whether or not Smith (or whoever) married (or whatever) of his own free-will, what reason have we for ever entertaining the sorts of arguments which engaged Hobbes and the churchman? And my answer is: none whatsoever; it depends on what you are after; it depends upon whether you are interested in doing history or metaphysics. But a meta-physical problem requires a meta-physical solution.”[125]

In summary, Flew’s arguments against the fourth reason for adopting this paradox do not succeed. Determinism makes plain that fatalism, properly construed, survives Flew’s challenge. Similarly, Flew’s argument for the conditional analysis of ability and Argument of the Paradigm Case remain controversial and have failed to persuade many philosophers that free will exists. Considering the weaknesses of these arguments, perhaps one should disagree with Flew and instead assert that “crime is a symptom of disease.”

Constructive Meta-Controllers

If “crime is a symptom of disease” then punishment has been its traditional medicine. The literature on free will suggests an alternative medicine. This literature is abundant with references to meta-controllers showing that certain agents are, or are not, morally responsible for their actions. These agents will be blamed, and punished, or praised, and rewarded. The literature on free will fails to suggest, however, that meta-control might present a better alternative to the tradition of rewarding and punishing agents. Meta-controllers can do more than prove the irrelevance of AP by failing to intervene (as in Frankfurt examples) or prove the nonexistence of free will by designing the lives of agents (whether the designer is secular, as in Derk Pereboom’s four-step argument, or divine, as in major monotheistic religions). Meta-controllers can do what each of these specific manipulators fails to do: they can provide a cure to the disease of immoral behavior*.

Frankfurt

Consider the Frankfurt controller. In order to show that agents can have moral responsibility without alternative possibilities, compatibilists suppose that the meta-controller fails to intervene. These examples do not emphasize, however, both that, if the agent was about to do something immoral, the meta-controller should intervene, and also that the meta-controller shares in the agent’s responsibility for whatever she or he does. So Frankfurt examples do not emphasize the fact that hard compatibilism cannot provide one with unique responsibility for one’s actions. Even in the actual scenario how one decides is not up to the agent or at least not just up to the agent. Frankfurt examples were intended to show, however, that moral responsibility survives this lack of alternative possibilities. So whatever moral responsibility survives must not be unique.

Pereboom

Similarly, Derk Pereboom considers meta-controllers, or “neuroscientists”, who program someone to kill another person:

“Case 2: Mr. Green is like an ordinary human being, except that he was created by neuroscientists, who, although they cannot control him directly, have programmed him to be a rational egoist, so that, in any circumstances like those in which he now finds himself, he is causally determined to undertake the reasons-responsive process and to possess the set of first and second-order desires that results in his killing Ms. Peacock.”[126]

Pereboom uses such situations in his four-step argument to establish incompatibilism which, together with less controversial assumption about indeterminism, shows that free will does not exist. This four-step argument has failed to persuade many compatibilists. Unlike Frankfurt examples, Pereboom’s four-step argument does emphasize how meta-controllers must share, or indeed usurp, a controllee’s responsibility. This is because Pereboom already denies the existence of free will.

The Future of Free Will

This suggests that Frankfurt controllers and Pereboom’s neurosurgeons can be used as an alternative for punishment. Instead of waiting for crimes to happen and then punishing people for wrongdoings, a Frankfurt controller could just watch these people and intervene before they commit a crime. Similarly, instead of having children in the traditional manner, Pereboom’s neurosurgeons could create agents who live perfect lives. In principle, Frankfurt controllers and Pereboom’s neurosurgeons could eliminate all crime, and therefore all punishment, from society. In this future, traditional beliefs in free will and moral responsibility, as well as traditional practices such as praising and blaming, are not non-existent so much as they are obsolete.

There are three important advantages of using meta-control as an alternative to punishment. For one, the deterrent effect of punishment is controversial. Punishment surely does deter some immoral behavior. But the extent to which it succeeds is unknown and may not be worth the pain inflicted. Secondly, punishment can reform people to only a modest extent at best. Hard compatibilists offer a world in which persons reward and punish each other without achieving significant reform. Punishment, in this world, serves as a check upon the excesses of a flawed human nature but can never perfect this nature. Meta-control, on the other hand, might create perfect agents. The final, and most important, advantage of using meta-control as an alternative to punishment is that meta-control has no pain cost. Lashing a person several times with a whip increases the disutility in the world; predicting the person’s crime prior to the event and intervening can be painless.

Indeed, one can generalize from the example of Frankfurt controllers and Pereboom’s neurosurgeons to the concept of a Lower Pain Cost Therapy (LPCT) with the following three criteria:

1. the LPCT has a lower pain cost than traditional punishment

2. the LPCT achieves the same deterrent effect as traditional punishment

3. the LPCT’s alters the person’s personal identity no more than traditional punishment does

LPCT’s need not involve convert neurosurgeons. Consider the example of a “prison moral pill” which alters the goals of its taker such that these goals becomes exactly like those the person would have had after a traditional prison sentence. There are two things to notice about this “prison moral pill”. For one, the pill is not just a Low Pain Cost Therapy but actually a Zero Pain Cost Therapy. This shows that one advantage of using LPCTs is that they can not just lower the pain associated with reform but actually eliminate it. The thing to notice about this “prison moral pill” is that the reforming it accomplishes, although having no pain cost, is still quite crude. Some people who take this pill, like some people who come out of prison, will still be bent upon committing new crimes. Suppose, however, that the LPCT is not just a “prison moral pill” but a “perfect agent moral pill” which alters the brain states of its takers such that crime becomes impossible for them. The “perfect agent moral pill” has a zero pain cost and reforms its takers far better than prison. Such LPCTs have an ethical advantage over traditional punishment.

Answers to Objections

This section has presented an alternative to the hard compatibilist worldview: immoral behavior* is a disease and meta-control, in the form of LPCTs, is the cure. The section will now anticipate and address several objections to this alternative: that this conception of disease is too broad; that meta-control does work as a general deterrent; that meta-control violates the dignity, autonomy, and personal identity of the controllee; that LPCTs are impossible and society will reject them; and that meta-control does not satisfy the demands of retributivism.

Too Broad a Conception of Disease

The most anticipated objection is that immoral behavior* is just not a disease. The strongest arguments against classifying immoral behavior* as a disease are that (i) immoral behavior* does not undermine a person’s capacities but may even express them and (ii) immoral behavior typically causes pain and disability, not in the subject, but in another. These objections suggest that, if this article has pressed the case for immoral behavior* as a disease too far, immoral behavior* nevertheless shares essential characteristics with diseases; in particular, immoral behavior* is typically not the subject’s fault.

Immoral Behavior* as an Adaptation

One significant objection to classifying immoral behavior* as a disease is that it does not seem to limit the capacities of people; indeed, it may even express their capacities more fully. Evolutionary psychologists have argued that a number of immoral behaviors evolved as adaptations because they increase fitness: “rape, homicide, infanticide, war, aggression, exploitation, infidelity, and deception.” [127] This is in distinct contrast to the Mutation Accumulation Theory and Antagonistic Pleiotropy Theory of Aging, according to both of which aging is not adaptive.[128] In this respect, the status of aging as a disease is less controversial than that of immoral behavior*. Immoral behavior* does not impair the phenotype of those who engage in it; it impair the phenotype of those who receive it.

Painfulness and Disability

The second principle objection to classifying immoral behavior* as a disease is that immoral behavior* typically causes pain and disability, not in those who engage in it, but rather in others. But this is not always true. Society tends to regard self-harm, such as self-mutilation, as immoral. Furthermore, some behaviors, which have been regarded as immoral, do not cause harm. Consider, for example, the immoral behavior of sodomy. In general, however, immoral behaviors do harm or disable others more than they harm or disable the subject.

Again, one response to this objection is just to deny that painfulness and disability is a criterion for a disease. For example, Caroline Whitbeck suggests the following criterion instead: “People wish to be able to prevent or terminate the process because it interferes with the bearer’s psycho-physiological capacity to do those things that people commonly wish and expect to be able to do.”[129] This definition is more conducive to classifying immoral behavior* as a disease because people (not necessarily the subject) do wish to prevent or terminate the process. The primary motivation for preventing or terminating immoral behavior* is to prevent harm or disability to others. But immoral behavior* also “interferes with the bearer’s psycho-physiological capacity to do those things that people commonly wish and expect to be able to do.” This might press the case for immoral behavior* as a disease too far, however.

Disease*

The previous arguments have attempted to show that immoral behavior* is a disease. Two objections to this view question this view, however, and the arguments may not have entirely succeeded. In that case, those who deny the existence of free will can rely upon a different concept: disease*. Disease* has the essential features of disease which those who deny the existence of free will would wish to preserve; disease* lacks, however, controversial attributes of disease which are not essential to free will denial. In particular, although disease* is (i) treatable in principle, (ii) requiring intervention and (iii) ordinarily not the subject’s fault, disease* (i) may not be adaptive and (ii) the harm of disability it causes may be to others and not just the subject. Disease, on this view, is a subset of disease*, where both are treatable in principle, require intervention, and are ordinarily not the subject’s fault, but only the former is non-adaptive and necessarily harms or disables the victim. Those who deny the existence of free will might grant that immoral behavior* is not exactly a disease while still wishing to say that it is sufficiently like a disease to be a disease*.

General Deterrence

One might also object that LPCTs have no general deterrence function. Philosophers distinguish between specific and general deterrence. Specific deterrence disinclines an offender from committing another crime. General deterrence disinclines the general popular from committing crimes. Although the specific deterrent effect of traditional punishment is controversial, it probably does work as a general deterrent. LPCTs might work in the opposite way: although (by definition) they would have a specific deterrent effect, they would have almost no general deterrent effect. A member of the general population might say “I am going to kill this innocent citizen and the only bad thing that will happen to me is that doctors will change my brain to make me a nicer person.”

This objection illustrates an unavoidable cost. To the extent that replacing punishment with LPCTs will cause a loss in general deterrence, there may be an increase in crime. But this cost may be compensated by other factors. For one, the large increase in welfare caused by the lack of punishment may compensate for the increase in crime. Secondly, society may wish to take preventative steps before members of the general population commit a crime. For example, if the pain-cost is sufficiently low then each citizen might receive a LPCT just to be safe. If these measures are not adequate compensation for the loss in general deterrence, however, then resorting to a modest amount of punishment may be necessary.[130]

Violations

A second objection to the alternative sketched here, whereby immoral behavior* is a disease* and LPCTs are the cure, is that this alternative involves the violation of personal rights. These rights include the right to dignity, autonomy, and to one’s identity. This section will address these objections in turn. In particular, the argument will be that this alternative violates none of these rights more than traditional punishment does.

Dignity

One might object to using LPCTs instead of punishment because doing so violates one’s dignity. Traditional punishment might already violate, however, the dignity of the punished and so the objection against using LPCTs must be more than this. Indeed, traditional punishment violates the dignity of those punished to the extent that it causes them suffering. If a LPCT causes little or no suffering then it is not clear how that therapy violates the subject’s dignity.

There is a Kantian objection that free will denial will undermine one’s dignity. Derk Pereboom addresses this objection in his Living Without Free Will. He concludes, however, that:

“If hard incompatibilism is true, then agents are not morally responsible for setting ends and choosing means, for formulating principles and making commitments to them. Nevertheless, the capacities for these activities can remain intact.”[131]

Later Pereboom address Herbert Morris’ novel objection to the use of LPCTs as an alternative to punishment:

“Morris adduces Kanian reasons in his argument against policies for criminal therapeutic rehabilitation. In his view, among the human qualities we value most is the capacity to regulate actions autonomously and rationally. The problem for typical forms of therapy proposed for altering criminal tendencies is that they circumvent, rather than address, these capacities. For example, consider the Ludovico method, made famous by Anthony Burgess’s book and Stanley Kubrick’s film A Clockwork Orange. Alex, a violent criminal, is injected with a drug that makes him nauseous while at the same time he is made to watch films depicting the kind of violence to which he is disposed. The goal of the method is eliminating the violent behavior by generating an association between violence and nausea. Morris’s objection to therapy of this sort is that the criminal is not changed by being presented with reasons for altering his behavior that he would autonomously and rationally accept. As an alternative, Morris argues for treatment that presents the criminal with a conception of the good that he could accept in this way. As he conceives it, this treatment must involve punishment, for by this means it would communicate to the criminal how his wrongdoing had adversely affected himself and others, and the appropriateness of guilt and repentance.”[132]

Pereboom notes, however, that there are two problems with Morris’ argument. For one, LPCTs might actually enhance or enable an agent’s rationality and autonomy. So such therapies may also enhance a subject’s dignity. The second objection Pereboom makes to Morris’ argument is that it violates consequentialist intuitions. Suppose, Pereboom asks, that there are two possible responses to a criminal: “indefinite confinement” and “behavioristic therapy,” such as the Ludovico method, which enhances the criminals’ reasons-responsiveness. Ruling out the Ludovico method is just contrary to widespread consequentialist intuitions.

Conservative thinker Francis Fukuyama suggests another reason why LPCTs might undermine human dignity:

“Denial of the concept of human dignity—that is, of the idea that there is something unique about the human race that entitles every member of the species to a higher moral status than the rest of the natural world—leads us down a very perilous path. We may be compelled ultimately to take this path, but we should do so only with our eyes open. Nietzsche is a much better guide to what lies down that road than the legions of bioethicists and casual academic Darwinians that today are prone to give us moral advice on this subject.”[133]

Fukuyama expresses concern that posthumans might necessarily reduce the rights of others (perhaps infants or the mentally handicapped). There is no reason, however, to think that this will necessarily be the case. Indeed, one might grant equal rights not just to infants and the mentally handicapped but also posthumans. Nick Bostrom responds to Fukuyama’s concern by arguing that the “moral circle” should expand to include those made of either “flesh or silicon”:

“The underlying intuition seems to be that instead of the famed ‘expanding moral circle,’ what we have is more like an oval, whose shape we can change but whose area must remain constant. Thankfully, this purported conservation law of moral recognition lacks empirical support. The set of individuals accorded full moral status by Western societies has actually increased, to include men without property or noble decent, women, and non-white peoples. It would seem feasible to extend this set further to include future posthumans, or, for that matter, some of the higher primates or human-animal chimaeras, should such be created—and to do so without causing any compensating shrinkage in another direction. (The moral status of problematic borderline cases, such as fetuses or late-stage Alzheimer patients, or the brain dead, should perhaps be decided separately from the issue of technologically modified humans or novel artificial life forms.) Our own role in this process need not be that of passive bystanders. We can work to create more inclusive social structures that accord appropriate moral recognition and legal rights to all who need them, be they male or female, black or white, flesh or silicon.”[134]

Autonomy

One might also claim that LPCTs violate the autonomy of a person. But, again, traditional punishment, such as prison, already violates the autonomy of a person and so one would need to show more than this. One might argue, for example, that LPCTs violate the autonomy of a person if they render that person into a perfect agent for whom immoral behavior is impossible. But LPCTs cannot reduce or enhance the free will of their subjects; perfect agents lack free will just as much as amoeba. Immoral behavior may be impossible for perfect agents but perfect lives are impossible for imperfect agents.

Indeed, there is no good reason to desire the ability to act immorally. This ability might have a superficial appeal. For example, if one had the ability to kill innocent people then this may prove useful if terrorist threatens to destroy the world unless one does kill such an innocent person. From this consequentialist perspective, however, this act is not immoral; it is an otherwise immoral act that becomes moral in context. But LPCTs would not threaten one’s ability to act morally in such situations. For example, LPCTs would threaten one’s ability to kill innocent people outside of this greater context. There is no reason, however, for one to desire this ability.

Personal Identity

A third violation to which one might object is the violation of personal identity. By altering one’s propensity towards immoral behavior, LPCTs seem to alter what kind of person one is. But, again, traditional punishments already alter one’s personal identity to this extent so one would need to show more than this. Convicts do come out of prison changed: they may be meaner (and perhaps less like to commit another crime). Indeed, the third criterion for a LPCT stipulates that it cannot alter a person’s personal identity more than traditional punishment does.

Traditional punishment cannot make someone a perfect agent, however. In that case, we can consider LPCT*s which do violate one’s personal identity. Both traditional punishment and LPCT*s might violate one’s personal identity along a spectrum. Traditional punishment makes a modest change in one’s personal identity to accomplish a modest amount of reform. By the same logic, LPCT*s which make agents perfect significantly alter one’s personal identity to accomplish a significant amount of reform. In any case, one wonders what is so precious about one’s imperfect identity.

Dennett

Daniel Dennett presents two novel arguments against the position this article defends. According to the first objection, LPCTs cannot preserve the “informedness, flexibility, and open-mindedness” of the agent and so are impossible. According to the second objection, society will refuse to use LPCTs and therefore preserve the sense that people have free will.

Impossibility

The following passage from Daniel Dennett’s Freedom Evolves explains his first objection. In context, Dennett has been arguing (against Mele) that brainwashing a person might undermine that person’s autonomy just to the extent the person remains uninformed about this intervention. He seems to go further here, however, and suggest that such autonomy preserving brainwashing is impossible in principle:

“The genuinely autonomous agent is rational, self-controlled, and not wildly misinformed. The intuitive repugnance we feel in ‘morality pills’ and ‘brainwashing,’ in contrast to good old-fashioned moral education, is perhaps due, then, to a dim appreciation of the utter impossibility of there being any such shortcut treatments that could actually preserve the informedness, flexibility, and open-mindedness that, in our experience, depends on a sound education.”[135]

Dennett continues by arguing that one’s autonomy might remain intact so long as the person knowingly uses the LPCT:

“I cannot see that knowingly taking a pill to improve one’s self-control is any more subversive to one’s autonomy than knowingly fostering a modest amount of self-deception about one’s powers. If you can knowingly manipulate yourself in these ways, as a consenting adult, and endorse the effects both prospectively and retrospectively, this is a fairly good test of whether you can justifiably manipulate your children in the same fashion.”[136]

Here Dennett seems to suggest that LPCTs are possible but that one must use them knowingly in order to preserve one’s autonomy. It is not clear, however, that “informedness, flexibility, and open-mindedness” are worth the pain cost that punishment inflicts. Suppose, however, that Dennett continues to insist that “moral pills” and “brain washing” cannot accomplish what “good old-fashioned moral education” can. But someone as committed to materialism as Dennett should recognize that such LPCTs are possible: just consider the LPCT that alters the subject’s brain to reflect all and only those changes to the subject’s goals that “good old-fashioned moral education” would.

The impact of Dennett’s argument upon the position of this article is complicated by the fact that he is defending the autonomy of people free from government intervention, whereas this article is considering those who may already justify such intervention. So Dennett may already, and perhaps justly, consider the autonomy of these latter agents to be compromised. In either case, despite Dennett’s suggestion to the contrary, LPCTs remain possible in principle.

There remains a rich irony in Dennett’s position on “moral pills” and “brainwashing.” Dennett conceives of Freedom Evolves as defending, not traditional free will, but a surrogate which has all of the same virtues. Consider, again, the following excerpt:

Dennett is eager to endorse the meaning of his naturalized “free will” as a surrogate for the traditional meaning of free will but refuses to endorse the use of LPCTs as an artificial replacement for punishment. This article has already shown that a hard compatibilist understanding of “free will” cannot “play all of the valuable roles free will has traditionally been invoked to play.” But LPCTs can play all of the valuable roles punishment has traditionally been invoked to play—without causing needless suffering.

A Medicalized Society

This objection leads Dennett to anticipate not just free will’s demise but also its replacement. He writes:

“The anxious mantra returns: ‘But where will it all end?' Aren't we headed toward a 100 percent ‘medicalized’ society in which nobody is responsible, and everybody is a victim of one unfortunate feature of their background or another (nature or nurture)? No, we are not, because there are forces—not mysterious metaphysical forces, but readily explainable social and political forces—that oppose this trend, and they are of the same sort, really as the forces that prevent the driving age from rising to, say, thirty! People want to be held accountable. The benefits that accrue to one who is a citizen in good standing in a free society are so widely and deeply appreciated that there is always a potent presumption in favor of their inclusion. Blame is the price we pay for credit, and we pay it gladly under most circumstances. We pay dearly, accepting punishment and public humiliation for a chance to get back in the game after we have been caught out in some transgression. And so the best strategy for holding the line against creeping exculpation is clear: Protect and enhance the value of the games one gets to play if one is a citizen in good standing. It is erosion of these benefits, not the onward march of the human and biological sciences, that would threaten the social equilibrium.”[138]

There are many things to note about this paragraph. For one, Dennett does not make a normative argument. He does not argue against medicalizing society (although he clearly wishes to avoid that). Instead, Dennett predicts that society will never medicalize itself entirely. To support his prediction, he cites forces working against this medicalization: people want to be held responsible and a Community of Responsibility has more benefits than a medicalized society.

Even if Dennett’s prediction were correct, his normative assumption might still be wrong. That the world refuses to medicalize itself might be a mistake. But even Dennett’s prediction is dubious. It evinces a failure of imagination. Sure, people want to be held responsible now. This is convenient because real LPCTs do yet exist. Indeed, punishment may be close to the lowest pain-cost therapy available today. In a distant or hypothetical future, however, society may no longer wish to be held responsible. It may wish to be treated. The claim that a Community of Responsibility has greater benefits than a medicalized society—and that society can ensure that the balance always remains in the Community of Responsibility’s favor—is just false. Such a Community of Responsibility will always have the staggering pain which a medicalized society can avoid: the systematic punishment of criminals. To compensate for this atrocity, there are no clear benefits which a medicalized society cannot also offer.

Retributivism

This section will address one final objection to the alternative sketched here. According to this objection using LPCTs as a cure for immoral behavior does not satisfy the demands of retributivism. Retributivism specifies that “punishment of a wrongdoer is justified for the reason that he deserves something bad to happen to him—pain, deprivation or death, for example—just because he has done wrong.”[139] Such punishment may be justified even when it furthers no consequentialist end.[140] Punishment, on this view, is punishment for its own sake.

One immediate problem with retributivism is that it seems not unlike sadism.[141] Both recommend actions which:

1. give the actor pleasure

2. achieve this pleasure through causing the subject suffering

3. achieve no other goal than this pleasure

The defenders of retributivism might respond to this analogy by trying to distinguish between vengeance and retribution. For example, retribution might be more limited in its severity than vengeance. This is a tenuous distinction, however, upon which to hinge one’s endorsement of retributivism.

The retributivist may also adopt the mitigating strategy of denying that an action may be right no matter what the consequences are. The following passage by Bernard Williams is representative of this view:

“This is not at all to say that the alternative to consequentialism is that one has to accept that there are some actions which one should always do, or again some which one should never do, whatever the consequences: this is a much stronger position than any involved, as I have defined the issues, in the denial of consequentialism. All that is involved, on the present account, in the denial of consequentialism, is that with respect to some type of action, there are some situations in which that would be the right thing to do, even though the state of affairs produced by one’s doing that would be worse than some other state of affairs accessible to one.”[142]

Similarly, James Montmarquet writes:

“I believe it is desirable, then, for a deontological view not to be committed (as mine is not) to the very strong view that killing the innocent is wrong, no matter what the consequences. For such a commitment exceeds the bounds of the conventional moral norms that deontological views aim to systematize. Wisdom suggests that we leave open the theoretical possibility that killing in such cases is justified, and move on to cases about which we have more determinate intuitions.”[143]

So retributivists are only retributivists up to a point. To be more precise, suppose that taking an action A1, such as punishing, which will result in a world W1, instead of taking a different action A2, which will result in a different world W2, even if W2 is a better state of affairs. Suppose further that W1 has value (in terms of utility, rights, or whatever concerns the retributivist) V1, whereas W2 has value V2. Let V3 be the difference between V2 and V1:

V2 – V1 = V3

Deontological theorists seem willing to advocate taking the action A1, instead of taking action A2, so long as V3 does not exceed a certain threshold T. For example, such a theorist might be willing to kill an innocent person to save the lives of two million people—but not just two.

0 < V3 < T à A1

0 < T < V3 à A2

Here is the practical problem with the retributivists objection to using LPCTs as a cure for immoral behavior*: as the pain cost of LPCTs becomes arbitrarily low, and the reformative function of LPCTs becomes arbitrarily effective, the value of V3 becomes arbitrarily large, and so the retributivist endorsement of A1 will become increasingly untenable. V3 is the benefit of free will denial. Retributivists will be unable to look upon two potential worlds—heaven and hell—and choose the latter. The prospect of LPCTs, like the prospect of meta-controllers, will have hard compatibilists concede that agents might not be worthy of blame, or punishment, despite being morally responsible for immoral behavior. Such hard compatibilists should find it telling that Antony Flew, who refused to classify crime as a disease, was not a retributivist:

“Surely we ought to interpret the notorious indifference of so many both of our popular Sunday journalists and of our learned and leading judges to the hard statistical data on the actual deterrent efficacy of floggist: not merely—with Dr. Friedlander—as the result of deplorable unconscious moties; but also at least partly as a matter of an equally deplorable—to Dr. Friedlander and the present writer—ethical convinction that merited fitting retribution must be inflicted, whether or not this is the most effective deterrent to others, or is necessary to or even compatible with the re-education of the offender himself.”[144]

There is a final remark to make about this future. If the argument of the previous section was correct, then the slavery of perfect agents is especially conspicuous. According to the argument of this section, however, society is morally obligated to using LPCTs to perfect the imperfect agents in this world. The upshot is that as these agents perfect themselves, like butterflies exiting their chrysalises, they will also shed their belief in free will. If, as Thomas More noted, it is vain to boast of making thieves and then punishing them, so too will it be vain to make saints and praise them.[145]

Three Kinds of Self-Skepticism

There is a natural harmony between the three views defended here: consequentialism, transhumanism, and free will denial. Unlike hard compatibilism, each of these views involves approaching the self with sufficient skepticism. For consequentialism, this implies aliening persons from their projects and undermining their integrity. For transhumanism, this implies a willingness to criticize human nature and an eagerness to revise it. For free will denial, this implies recognizing that one’s life is, in Saul Smilansky’s lovely phrase, “the unfolding of the given.”[146]

Consequentialism

Although philosophers often assume that free will denial implies consequentialism, one may wonder why this is so. For example, Michael Slote has argued that virtue ethics is consistent with free will denial.[147] Derk Pereboom, agreeing with Slote, doubts that the non-existence of free will implies consequentialism:

“The reason for this is that the metaphysical bases for non-consequentialist positions in general, insofar as they have been developed, do not clearly involve an essential appeal to notions of freedom unavailable to the hard incompatibilist. For example, absolutist restrictions on consequentialist principles do not entail that we are free in the sense required for moral responsibility or that we can be praiseworthy and blameworthy.”[148]

If the retributivist’s claim is properly construed then Pereboom’s conclusion follows. One doubts, however, whether such a retributivist would prescribe rules to govern non-autonomous agents. Consider the rule “agents ought not to lie.” This rule may be compatible with free will denial. A retributivist may be inclined, however, to frame this rule in terms of autonomous agents: “autonomous agents ought not to lie.” Formulated this way, the rule loses all prescriptive force in a world without autonomous agents.

Pereboom continues by examining Kant’s two formulations of the Categorical Imperative and concluding that it is compatible with free will denial. Again, these formulations of the Categorical Imperative do not limit its application to autonomous agent. This Categorical Imperative may be compatible with free will denial just as the consequentialist rule “act so as to produce the best state affairs” is. One suspects, however, that retributivists are assuming that this rule only applies to autonomous agents. If they are making this assumption then the Categorical Imperative loses its force in a world without autonomous agents; the consequentialist rule, however, may not have any such restriction.

To determine whether retributivism does tend to stipulate that its rules only apply to autonomous agents, it may be helpful to inquire into the beliefs of prominent defenders of that view. For example, Galen Strawson notes that, according to Immanuel Kant, knowledge of the moral law is proof that the agent is autonomous.[149] Similarly, Strawson cites Kant’s writings to show that he believed in the radical sort of freedom the existence of which this article denies:

“Since he was committed to belief in radical moral responsibility, Kant held that such self-creation does indeed take place, and wrote accordingly of ‘man’s character, which he himself creates’ and of ‘knowledge of oneself as a person who… is his own originator.’”[150]

Kant—perhaps the quintessential non-consequentialist—believed that any agent who knows the moral law must also possess free will of a radical kind. So Kant may have stipulated his Categorical Imperative without limiting its application to autonomous agents because he felt that any agent who knew of the rule must be autonomous. In contrast, consider the writings of prominent consequentialists. Such consequentialists either denied the existence of free will, such as William Godwin, or defended a progressive sort of compatibilism. For example, Henry Sidgwick criticized Kant’s view on free will to the extent that it committed the mistakes of libertarianism:

“But it is clear that if we say that a man is a ‘free’ agent in so far as he acts rationally, we cannot also say—in the same sense—that it is by his own ‘free’ choice that he acts irrationally, when he does so act; and it is this latter proposition which Libertarians generally have been concerned to maintain.”[151]

Similarly, although John Stuart Mill was a famous compatibilist, he also wrote in Utilitarianism:

“As a refuge from the last of the three, men imagined what they called the freedom of the will; fancying that they could not justify punishing a man whose will is in a thoroughly hateful state, unless it be supposed to have come into that state through no influence of anterior circumstances.”[152]

In general, consequentialists may be more likely to advocate their view despite the non-existence of free will and moral responsibility. Historically, consequentialism seems to have been conducive to the free will denial and progressive compatibilism. Non-consequentialists such as Kant, however, may be more inclined to restrict the application of their rules to free agents.[153]

Pereboom

One might question whether Pereboom is a consequentialist. Some of his writing suggests that he is not. For example, Pereboom takes pains to show that free will denial does not limit one to consequentialism and he argues against a utilitarian theory of punishment. Furthermore, although utilitarianism is just a species of consequentialism, Pereboom employs arguments against the former which might apply with equal force to the latter.

Pereboom cites three arguments against utilitarianism: it might justify punishments that are otherwise too severe, it might justify punishing the innocent and it might justify using people in an otherwise unjust manner.[154] These objections lead Pereboom to consider a self-defense rationale for punishment which he further refines into a justification for quarantining. Punishment, on Pereboom’s view, is never or almost never justified.

One might question, however, whether Pereboom’s arguments against utilitarianism, and consequentialism more broadly, succeed. For example, Pereboom anticipates John Rawls’ argument that punishing the innocent might not bring about the best state of affairs. Too severe punishments might also not bring about the best state of affairs. Pereboom’s only answer to Rawls’ objection, however, is that “this practice would seem to be more deeply wrong than can be accounted for by the utilitarian reasons he presents.”[155] But if punishing innocents is intended to prevent a sufficiently trivial evil, Rawls’ objection is plausible; if punishing innocents is intended to prevent a sufficiently severe evil, the unwillingness to do so may be immoral. Similarly, using agents in an otherwise inappropriate manner may be necessary to prevent a greater evil. Furthermore, one might detect a tension here between Pereboom’s unwilling to use agents and his willingness to use therapies which circumvent an agent’s autonomy and rationality. Those who administer such therapies may be using the agent for the purposes of the agent and themselves.

There is another objection to Pereboom’s argument against utilitarianism, and consequentialism more broadly: the distinction between quarantining and punishment is artificial. I could not find an explicit definition of the punishment concept that Pereboom uses in Living Without Free Will. Pereboom does note that “[t]he central cases of punishment imposes on an offender serious harm, such as long-term loss of liberty by confinement in the sorts of prisons we have in our society, significant physical or psychological pain, or death, because he has done wrong.”[156] But quarantine does impose a serious harm on an offender: the long-term loss of liberty by confinement. The harm is so grave that those who are quarantined unjustly seek compensation. Indeed, Pereboom seems to prefer the punishment of quarantining for largely utilitarian reasons.

One way to forge this distinction between quarantining and punishment is to focus upon intent. Perhaps quarantining is motivated by concern about deterrence whereas punishment involves an element of retribution. But in that case one might consider graver harms than quarantining to not be punishment. Indeed, one might justify the death penalty as a deterrent alone. On that view, the death penalty might not be punishment so much as it is a necessary evil. This distinction helps to explain the puzzlement which Daniel Dennett expresses towards those who deny the existence of free will:

“Hard determinism: Determinism is true, so we don’t have free will. Hard-headed scientific types sometimes proclaim their acceptance of this position, even declaring it a no-brainer. Many of them would add: And if determinism is false, we still don’t have free will—we don’t have free will in any case; it’s an incoherent concept. But they typically excuse themselves from exploring the question of how they then justify the often strongly held moral convictions that continue to guide their lives.”[157]

Consider executing an innocent bystander to prevent a thermonuclear war in which billions of people will die. Now consider an observer who does not know the context of the execution. The observer only sees one person killing another for no apparent reason. This looks just like punishment; but is it? The puzzlement which hard compatibilists such as Dennett express toward the behavior of those who deny the existence of free will may be not unlike that expressed by this observer. The purpose of this article is, in part, to fill the gap that Dennett notes in exploring the question of how those who deny the existence of free will “then justify the often strongly held moral convictions that continue to guide their lives.” Once Dennett understands the consequentialist motivation for “punishing” criminals, he may no long feel that those who deny the existence of free will are being inconsistent. Similarly, once Pereboom understands the consequentialist motivation for “punishing” criminals, he may begin to either classify quarantining as punishment or grant that graver harms than quarantining may be appropriate, in principle, for criminals. Nevertheless, Pereboom’s emphasis upon quarantining fits well with this article’s conclusion that immoral behavior* is a disease*.[158]

Smilansky

Saul Smilansky, another eloquent spokesperson for free will denial, presents one of the other few explorations of free will and meta-ethics.[159] On Smilansky’s view, one can distinguish between two types of compatibilism: Effect Compatibilism and Control Compatibilism. Effect Compatibilism holds that punishment is proper when it is “effective, for example, as a deterrent.”[160] The primary difference between Effect Compatibilism and Control Compatibilism is that the latter respects “[t]he elementary ethical conception that takes as its focus the necessity of considering free will, in some sense[.][161]” Smilansky calls this elementary concept the “Core Conception.” Effect Compatibilism does not first scrutinize the freedom of agents and then assign punishment. Rather, Effect Compatibilism first considers whether punishment would be effective, and then assigns punishment. Effect Compatibilism is largely similar to my own view.

Smilansky rejects Effect Compatibilism and prefers Control Compatibilism. He prefers the latter for two reasons: Effect Compatibilism (Smilansky argues) would have dire consequences and Effect Compatibilism does not respect the Core Conception. But neither of these objections is convincing.

Smilansky’s conclusion that Effect Compatibilism would have dire consequences relies, in large part, upon his selection of utilitarianism as the test for whether punishment would be effective. According to this test, punishment would be effective just in case it maximized future utility. Smilansky objects that such a system would justify punishing the innocent. But punishing the innocent may not always be wrong. Few would argue that one should not punish one innocent person in order to avoid a thermonuclear war in which billions die. Furthermore, to the extent that punishing the innocent is not worth future utility, then the problem is not Effect Compatibilism per se, but its utilitarian character. There is irony in criticizing consequentialism because of its consequences; one needs to select a different ethical system and not a different meta-ethical system.

Smilansky also rejects Effect Compatibilism because it does not do justice to the Core Conception. Only Control Compatibilism can do that. But, on Smilansky’s definition of “punishment,” the Core Conception does not deserve strict respect. Smilansky, like Pereboom, does not stipulate that punishment implies desert. He uses the term in the context of serious harms regardless of desert. So, on Smilansky’s definition of “punishment,” the Core Conception specifies that the allocation of serious harms should consider “up to us-ness.” But on this reading, Smilansky’s objection about the Core Conception translates into his original objection about punishing innocents. That objection fails, however, because punishing innocents is not always wrong. Furthermore, if those who allocate punishment find that, despite their ethical system’s recommendation, the punishment of innocents is ever not worth the apparent advantages, then they should choose a different ethical system.[162]

A more natural reading of “punishment” provides a more elegant reply to Smilansky’s objection about the Core Conception. If punishment is just serious harm, then chemotherapy might be punishment. But this seems intuitively wrong. Punishment implies not just serious harm but the serious harm of those who deserve it. On this more natural reading of punishment, Effect Compatibilism might still respect the Core Conception. Rigorous application of the Core Conception would deny that anyone ought to be punished because, ultimately, desert is impossible. Effect Compatibilism implies that punishments should only be given to those who deserve them. But society might satisfy the Core Conception by systematically allocating serious harms without making any attribution of blame. In this world, society harms criminals but it does not punish them.

Intergrity

There is another aspect to the intimate relationship between consequentialism and free will denial. Bernard Williams famously accused a form of consequentialism, act utilitarianism, of alienating persons from their projects and undermining their integrity:

“It is absurd to demand of such a man, when the sums come in from the utility network which the projects of others have in part determined, that he should just step aside from his own project and decision and acknowledge the decision which utilitarian calculation requires. It is to alienate him in a real sense from his actions and the source of his action in his own convictions. It is make him into a channel between the input of everyone’s projects, including his own, and an output of optimific decision; but this is to neglect the extent to which his actions and his decisions have to be seen as the actions and decisions which flow from the projects and attitudes with which he is most closely identified. It is thus, in the most literal sense, an attack on his integrity.”[163]