Why All Frankfurt-Style Examples Fail to Help Compatibilism

 “Why All Frankfurt-Style Examples Fail to Help Compatibilism” James W. Sperman Section I Introduction: Incompatibilists contend that free will requires alternate possibilities1 . Then, since determinism2 doesn’t allow for alternate possibilities, they conclude that free will and determinism are incompatible. One nifty way for compatibilists to challenge this conclusion is to point to the following argument. If moral responsibility requires free will, and if free will requires alternate possibilities, then moral responsibility requires alternate possibilities. But if moral responsibility doesn’t actually require alternate possibilities, then either moral responsibility doesn’t require free will or free will doesn’t require alternate responsibilities. And since it is generally agreed that moral responsibility requires free will3 , it must be the case that free will does not really require alternate possibilities. All Frankfurt-style examples4 are of this sort; that is, they attempt to defend compatibilism by showing that moral responsibility does not require alternate possibilities. If they succeed, then, as a logical consequence, free will and determinism will be compatible via the argument in the preceding paragraph or one like it. If they fail, however, then compatibilism will not be helped by them. 4 See “Alternate Possibilities and Moral Responsibility” for Harry Frankfurt’s original example 3 John Martin Fischer is one notable dissenter, but in my next paper on free will I intend to show precisely why his position is untenable. 2 I will argue from within the assumption that some brand of physicalism is true. Therefore, “determinism” will mean “physicalist determinism” throughout this paper. 1 I owe special thanks to Robert Kane, whose pellucid book A Contemporary Introduction to Free Will not only framed the Frankfurt example issue, but served as my invaluable guide through the topic of free will. 2 My aim in this paper is to show that no Frankfurt-style examples can help compatibilism. My argument is this: 1) In order to help compatibilism, a Frankfurt-style argument must show that moral responsibility does not require alternate possibilities. 2) In order to show that moral responsibility does not require alternate possibilities, a Frankfurt-style argument must disprove Strong PAP (SPAP: if a person A is morally responsible for an action Z, then A had alternate possibilities for Z or for a relevant5 antecedent event to Z). 3) No actual or possible Frankfurt-style example can disprove SPAP. 4) Therefore, no Frankfurt-style example can show that moral responsibility does not require alternate possibilities. (2,3) 5) Therefore, no Frankfurt-style example can help compatibilism. (1,4) The flow of this paper will be as follows. I will end this section by presenting a generalization of traditional Frankfurt-style examples. I will then advert to the task of showing my main argument to be sound. Since I have already accounted for the truth of the first premise, and since it is not a point of contention, my efforts in Sections II and III will center on supporting the truth of the second and third premises, respectively. If my arguments in these sections are successful, then my main argument will be sound and all Frankfurt-style examples will be shown to fail in their attempts to help compatibilism. Traditional Frankfurt-style examples: 5 In this context, an event E is relevant if and only if A’s moral responsibility for Z is wholly or partly engendered by E. 3 A generalized version of a traditional Frankfurt-style thought experiment goes like this. Person A is considering whether to do a Bad Thing. Agent C has the following power over A: if A is about to choose to not do the Bad Thing, then C will intervene and force A to choose to do the Bad Thing; however, if A chooses on his own to do the Bad Thing, then C will remain inert. Frankfurt’s argument is this. If A chooses to do the Bad Thing without any intervention by C, then A is morally responsible even though he could not have chosen otherwise; however if C intervenes and forces A to choose to do the Bad Thing, then A is not morally responsible. And why is this? Because in the former case A chooses on his own to do the Bad Thing, whereas in the latter case it is really C that makes the decision (by intervening and overruling A’s attempt to choose otherwise). So either way A will end up choosing to do the Bad Thing, but his moral responsibility is contingent because it hinges on whether C is forced to intervene or not. New Frankfurt-style examples of greater complexity have been proposed that do not fit neatly into this generalization. These new examples attempt to overcome incompatibilist claims that the traditional examples fail in indeterministic scenarios. I will return to these new examples later in the paper and demonstrate that they are unsuccessful, but I wish to set them aside until their relevancy can best be shown. So for now, let us begin with just this generalized version of traditional Frankfurt-style examples. Section II 4 Realizing the importance of SPAP: To understand the importance of SPAP, we must first see how intuitions involving alternate possibilities enter into the picture. We can do this by noting that an intuitive tension is commonly felt between determinism and moral responsibility, and that this tension seems to arise from reasoning of the following type: 1) If determinism is true, then all of a person’s thoughts and actions were entirely and unalterably determined prior to his birth by P (P: all physical laws and the actual physical state of the universe at some time prior to his birth). 2) If all of a person’s thoughts and actions were entirely and unalterably determined before his birth, then, given P, and given that he can change nothing about P, he could never have done anything otherwise. 3) (AAP) If a person could never have done anything otherwise, then he could not be held morally responsible for anything he has done. 4) Therefore, if determinism is true, then no person could ever be morally responsible for anything he has done. The third premise seems to be the primary intuition for thinking that determinism and moral responsibility are incompatible. Let us term this intuition the Appeal to Alternate Possibilities: Appeal to Alternate Possibilities (AAP): if a person could never have done anything otherwise, then he could not be held morally responsible for anything he has done. When Harry Frankfurt sought to show that alternate possibilities are not necessary for moral responsibility, he coined his target the Principle of Alternate Possibilities (PAP) and characterized it in the following way: 5 Principle of Alternate Possibilities (PAP): “a person is morally responsible for what he has done only if he could have done otherwise.” 6 His strategy was to undermine our intuition of AAP by disproving PAP. This strategy, however, only succeeds if AAP implies PAP. It would be prudent, then, to analyze whether or not AAP actually does so. When we attempt this analysis, however, we notice that Frankfurt’s phrasing of PAP is ambiguous in that it lends itself to a range of interpretations. The extremes of these interpretations can be stated as follows: Very Weak PAP (VWPAP): if a person A is morally responsible for an action Z, then A could have done Z otherwise Very Strong PAP (VSPAP): if a person A is morally responsible for an action Z, then A could have done something7 otherwise There are, of course, a variety of interpretations that fall between these extremes; these will be interpretations in which A could have done something otherwise, but rather than “something” denoting general existential quantification, “something” will pertain to something specific8 . To determine the proper interpretation of PAP, we must look to the actual example put forth by Frankfurt. When we do, we see that two facts combine to make the proper interpretation clear. First, Frankfurt’s example is obviously not meant to disprove VSPAP, for the intuitive hook of the example is that the person could have done something otherwise: he could have forced the evil agent to intervene. Since this means that the example relies on the truth of VSPAP for its effectiveness, we can therefore infer that PAP—the principle Frankfurt is seeking to disprove—should not be interpreted as 8 Let’s call these Q-Principles, and note that they would be of the following form: if a person A is morally responsible for an action Z, then A could have done Q otherwise. 7 This is meant as “there exists some antecedent X such that X could have been done otherwise.” 6 This quote is taken from Frankfurt’s “Alternate Possibilities and Moral Responsibility,” which can be found on page 167 of Gary Watson’s Free Will anthology for the Oxford Readings in Philosophy, 2003. 6 VSPAP. Second, Frankfurt purports to disprove PAP with an example in which a person is morally responsible for a choice even though that choice could not have been avoided. It is important to note that there is no reference in his example to a need for any other antecedent “something” that could have been done otherwise; rather, being morally responsible for an act that could not have been done otherwise is presented as a sufficient condition for disproving PAP. This point allows us to eliminate from consideration the range of possible interpretations between VSPAP and VWPAP, since it is clear that none of these would be disproved by Frankfurt’s example9 . VWPAP, however, is disprovable by Frankfurt’s scenario. Therefore, since VWPAP is the only reasonable interpretation that can be disproved by Frankfurt’s scenario, we should conclude that the proper interpretation of PAP is VWPAP. Frankfurt succeeds in highlighting the moral intuition that a person’s moral responsibility for a given act is engendered not just by a particular act, but also by relevant antecedent events. I grant that by so doing he disproves VWPAP (which he terms PAP), but does this really advance him toward his goal of helping compatibilism? It seems that it does not, for if he must undermine AAP in order to help compatibilism, then since VWPAP is not implied by AAP, disproving VWPAP fails to undermine AAP. So what would disprove AAP? Well, disproving VSPAP certainly would, for it is the contrapositive of AAP10 . But I don’t think that compatibilists should have to disprove something as strong as VSPAP. My reasoning is this. In the deterministic scenario in which AAP arises, a person can, for any given act Z, do neither relevant antecedent acts 10 To be precise, VSPAP is the contrapositive of AAP that is then applied to one particular act Z. The point, however, is that disproving VSPAP would disprove AAP, since AAP entails VSPAP. 9 Q-Principles (if a person A is morally responsible for an action Z, then A could have done Q otherwise) can’t be disproved by a thought experiment that doesn’t falsify the consequent. So since Frankfurt’s thought experiment doesn’t disallow alternate possibilities for Q, it doesn’t disprove Q-Principles. 7 nor irrelevant antecedent acts otherwise. But by definition, the irrelevant antecedent acts are irrelevant. It seems clear, then, that AAP is stronger than it really needs to be to capture our moral qualms with a deterministic scenario. What we should do, then, is carve out the true moral intuition at work within AAP and separate it from the irrelevant chaff. If we do this, we are then left with the following root principle: Root AAP (RAAP): if for every act Z, a person A could never have done Z, or a relevant antecedent act to Z, otherwise, then A could never be held morally responsible for anything he has done. RAAP, then, is the root moral intuition at work when we consider deterministic scenarios, and AAP probably strikes us as intuitively true because it implies RAAP. In order to disprove AAP, a Frankfurt-style example would need to disprove VSPAP. But if my reasoning is correct, then a Frankfurt-style argument would not need to disprove AAP, but rather the weaker RAAP. Now the bad news for the Frankfurt camp is that disproving VWPAP doesn’t do any more good against RAAP than it did against AAP. The good news, however, is that instead of having to disprove VSPAP, a Frankfurt-style argument would only need to disprove a weaker principle that I’ll term Strong PAP: Strong PAP (SPAP): if a person A is morally responsible for an action Z, then A could have done Z, or a relevant antecedent act to Z, otherwise. Let’s pause now and take stock of where we are. We began by noting that many people have a strong intuition that moral responsibility is incompatible with determinism, and that this intuition is what a Frankfurt-style example would need to undermine. This intuition appears, in its rough form, as AAP (AAP: if a person could never have done anything otherwise, then he could not be held morally responsible for anything he has done). But AAP is excessively broad in that it exceeds the intuition at its root. This root 8 intuition is termed above as RAAP (RAAP: if for every act Z, a person A could never have done Z, or a relevant antecedent act to Z, otherwise, then A could never be held morally responsible for anything he has done). This demarcation is important in our current context because it means that Frankfurt-style examples need not undermine AAP in order to help compatibilism, but just the weaker RAAP instead. Now we found that Frankfurt’s PAP corresponds to what we have termed VWPAP, and that VWPAP will disprove neither AAP nor RAAP. Frankfurt-style examples must therefore disprove a more powerful principle than VWPAP (aka PAP). If such an example needed to undermine AAP in order to help compatibilism, then VSPAP would have to be disproved; however, since RAAP is the relevant root of AAP that must be undermined, I argue that a Frankfurt-style example only needs to disprove a weaker principle: SPAP. If I’m wrong in asserting that Frankfurt-style examples only need to disprove SPAP, then they’ll have to disprove the more powerful VSPAP. Of course, that potential problem is simply overkill if Frankfurt-style examples can’t even disprove SPAP. Section III Can any Frankfurt-style example show SPAP to be false? I claim that no Frankfurt-style example can show SPAP to be false, and my argument goes like this. In order for a Frankfurt-style example to show SPAP to be false, the example must present a scenario in which SPAP’s antecedent is true but its conclusion is false; that is, a person A is morally responsible for an action Z, but A had no alternate 9 possibilities for Z or for any relevant antecedent event to Z. Now A’s natural behavior11 in Frankfurt-style examples can only be presented as deterministic, indeterministic, or as a hodgepodge of both. If it is presented as indeterministic, then the example will be unable to show the consequent to be false when the antecedent is true. If it is presented as deterministic12 , then the example will be unable to show that the antecedent of SPAP is true. Finally, if it is presented as a hodgepodge, then the only way for it to be logically coherent is via a route that results in its collapse to indeterministic behavior. Therefore, since no option yields a scenario in which SPAP’s antecedent is true and its consequent false, no Frankfurt-style example can disprove SPAP. My goal in this section is to show that the above argument is sound. Let us work in reverse order, then, and begin by methodically analyzing the possible claim that A’s natural behavior is a hodgepodge of both deterministic and indeterministic behavior. If A’s natural behavior is a hodgepodge: The appeal of this option probably centers on a hope that events might only be indeterministic up to a certain point, either temporally or in terms of levels of reality, after which events follow deterministically. The upshot of such a scenario is that there may be some brain state that precedes A’s decision and that serves as an indicator for C that a certain decision will necessarily follow. If this scenario were possible, C would only need to have power over events up to and including the indicator, after which C 12 I will be thorough and address all options, though it is generally assumed that Frankfurt-style examples do not involve deterministic worlds. (As Mele and Robb state in “Rescuing Frankfurt-Style Cases,” “there is a significant dif erence between our scenarios and scenarios set in deterministic worlds: in ours, as is supposed to be the case in standard Frankfurt-style scenarios in general, the agent’s decisions lack deterministic causes.” Pg. 110) 11 A’s natural behavior is deterministic iff macro-level determinism is true, and is indeterministic iff macro-level indeterminism is true 10 could be assured that A would make the decision that C wanted him to make. The problem with this scenario is that it’s not logically possible. To see why, let’s consider each of the four routes that the hodgepodge option can take. One route is to claim that the micro-level is indeterministic and that the macro-level is deterministic. But this is impossible. As Glymour and Sabatés13 have recently shown, if the micro-level is indeterministic, then the macro-level must be indeterministic as well. A second route is to claim that the micro-level is deterministic and that the macro-level is indeterministic due to the supervenience relation being indeterministic. But if this is the case, then this route can be grouped with indeterminism, for no micro-level or macro-level state will be able to serve as a deterministic indicator of A’s behavior. A third route is to claim that deterministic and indeterministic entities exist concurrently on the micro-level. If this is the case, then it would either be indeterministic whether a deterministic micro-entity would interact with an indeterministic micro-level entity, or it would be deterministic whether an indeterministic micro-entity would interact with a deterministic micro-entity. But both of these options are contradictory nonsense. Therefore, it is impossible for deterministic and indeterministic entities to coexist on the micro-level. The final route is to claim that deterministic and indeterministic entities coexist on the macro-level. Now the third route shows us that the micro-level must be either all deterministic or all indeterministic. If the micro-level is indeterministic, then the first 13 See “Micro-Level Indeterminism and Macro-Level Determinism” by Bruce Glymour and Marcelo Sabatés, Proceedings of the Twentieth World Congress of Philosophy, Volume X: Philosophy of Science, Edited by Tian Yu Cao. 11 route shows us that the macro-level is necessarily indeterministic. If the micro-level is deterministic, then we have two options: if the supervenience relation is indeterministic, then the second route shows us that A’s behavior is entirely indeterministic; and if the supervenience relation is deterministic, then physicalism itself requires that the macro-level is entirely deterministic. Therefore, since each option results in the macro-level being either entirely deterministic or entirely indeterministic, this final route is logically impossible. So since the only hodgepodge option that is logically possible is one in which no marker on any level can serve as a deterministic indicator of A’s behavior, the only options that need to be considered for A’s natural behavior are the wholly deterministic and wholly indeterministic options. If A’s behavior is deterministic: The goal here will be to demonstrate that no Frankfurt-style example can show A to be morally responsible if A’s behavior is naturally deterministic. If I’m successful, then no such Frankfurt-style example can disprove SPAP, for no such example will be able to show SPAP’s antecedent to be true. Let’s first eliminate some needless clutter. If A’s natural behavior were deterministic, there would be absolutely no need for C or any counterpart to C 14 . This is because any scenario in which A would be judged morally responsible would be one in which A chooses what C wants A to choose. So since it would have been determined billions of years prior that A would choose on his own to do The Bad Thing, there would be no risk of A choosing otherwise and no chance that C would need to intervene. 14 For brevity’s sake, I will only mention C in the remainder of this section; but it should be understood that “C” represents C and all possible counterparts to C. 12 Therefore, since unnecessarily keeping C in the in the example would only muddle the picture and perhaps lead to confusion, he can and should be eliminated. The elimination of C represents a problem for compatibilists, however, for it seems to rob the example of its intuitive hook. For most folks, the intuitive hook is the sense that it is possible for A to trigger C’s intervention by trying to do otherwise. But C’s elimination makes it explicitly clear that it is impossible for A to trigger C’s intervention. And since the intuition of A’s moral responsibility is triggered by the sense that he actually could have at least tried to do otherwise, the intuitive power of the example would be lost. In place of the lost intuitive hook, Frankfurt-style examples would now feature question-begging. For if we assume that all of A’s actions and mental events were determined long before his birth, then we’re still faced with the following question: Is A morally responsible for his deliberations (and his ensuing decision and Bad Act) when, given the universe’s actual nature and past, he could not have deliberated otherwise? At this point, the pressure of a consequence-style incompatibilist argument would be crushing. Such an argument regarding moral responsibility would center on the claim that if our current actions are necessary consequences of events/states that we were not morally responsible for, then we are not morally responsible for our current actions. A simple version of this type of argument could run as follows: since A’s decision to do the Bad Thing follows as a necessary consequence of his initial brain state as a fetus15 , and since A is not morally responsible for his initial brain state as a fetus, then A is not morally responsible for his decision to do the Bad Thing. This argument has strong intuitive appeal and cannot be overcome without a non question-begging example to 15 In conjunction with physical law and the state of the universe at that time 13 undermine it. Therefore, since determinism-based Frankfurt-style examples would be reduced to question-begging, they could not overcome consequence-style arguments; and if they could not overcome these arguments, they could not show that A is morally responsible when he has no relevant alternate possibilities. Therefore, no Frankfurt-style example could falsify SPAP if A’s natural behavior is deterministic. If A’s natural behavior is indeterministic: If A’s natural behavior is indeterministic, then either it is possible for him to be morally responsible or it’s not. If it’s not possible, then the antecedent of SPAP will be false and SPAP will not be disproved. Therefore, in order to disprove SPAP, a scenario must be presented that concurrently shows that A is morally responsible and that A had no relevant alternate possibilities. So there are two questions we must ask. Do any existing Frankfurt-style examples provide such a scenario? And if not, is it even possible for a Frankfurt-style example to provide such a scenario? Let’s begin by looking at the generalized version of the traditional Frankfurt-style example. Now if A’s natural behavior is indeterministic, then his deliberations will be indeterministic until the very moment that A makes his decision about whether to do the Bad Thing. But if A’s deliberations are indeterministic right up to the point of the decision, then how could C infallibly predict A’s decision? If A’s deliberations are indeterministic, then there would be no surefire signs for C to look for that would indicate that A was about to decide to not do the Bad Thing. Therefore, it seems like the best C could do is to wait until A makes his decision, then immediately change the decision if it doesn’t match C’s wishes. But then A would in fact have alternate possibilities, because A would decide the instant before C would intervene. And if C intervenes prior to A’s 14 decision in order to eliminate the alternate possibilities, then A cannot be said to be morally responsible, for C would preemptively make the decision in place of A. New Frankfurt-style examples have been created to respond to this incompatibilist defense. One such response by Mele and Robb16 can be quickly generalized as follows. Imagine that a deterministic program P* has been loaded into A’s brain, and that P* will necessarily produce a desired Bad Decision17 at a particular time t unless A happens to arrive at the Bad Decision on his own. If A arrives at the Bad Decision on his own at t, then P* will become inert and A’s own efforts will be the direct cause of A’s decision; however, if A tries at t to make a Good Decision, or if A is still deliberating at t, then P* will actively produce the Bad Decision in A at t. So A will necessarily arrive at the Bad Decision, and the only difference will be whether he arrives at the decision on his own or due to P* commandeering his brain and actively producing the decision. This response seems inadequate in two glaring ways. First of all, it is logically incoherent to say both that the program P* is deterministic and that A’s deliberations are indeterministic. A corollary of the Glymour-Sabatés Implication is that the only way P* can be deterministic is if the micro-level is deterministic and if supervenience is a deterministic relation. Now if the micro-level and the supervenience relation are both deterministic, A’s own deliberations must be deterministic as well. But Mele and Robb claim that A’s deliberations are indeterministic. So as it stands, this response is incoherent and, as a result, fails to disprove SPAP. Compatibilists could respond by asserting that this incoherency could be resolved by changing P* from a deterministic program to a highly probable one. By doing so, 17 Mele and Robb use the decision to wrongfully steal a car in their essay. 16 Alfred Mele and David Robb, “Rescuing Frankfurt-style Cases,” Philosophical Review 107 (1998): 97-112 15 there would be a small percentage of scenarios in which P* fails and A has alternate possibilities; but this would be unimportant. The relevant point is instead that A would usually have no alternate possibilities, and that this would be close enough to determinism for the purpose of disproving SPAP. I’m prepared to grant that compatibilists could salvage the coherency of the P* example in the preceding manner, but I contend that this move will not render their example effective. For even if we were to set the incoherence issue (and its dubious solution) aside, it seems clear that A would still have alternate possibilities: he could make the Bad Decision on his own or he could force P* to have to complete its program and make the decision for him. If A refrains from making the Bad Decision on his own and thereby forces P* to enact it, then A will be free of moral responsibility and SPAP will survive. And if A engenders moral responsibility due to making the Bad Decision on his own, then SPAP still escapes since A had the relevant alternate possibility of forcing P* to make the decision. At this time I must pause to make absolutely certain that there is no confusion about my position. Assuming that indeterminism is true and that indeterministic free will is coherent18 , I contend that A is morally responsible for the Bad Decision if A makes the decision on his own, regardless of whether he could have avoided reaching that decision. But under the same assumptions, if A tries to do otherwise and something beyond his control overrides his attempt and forces him to make the Bad Decision, then he is not morally responsible for that decision. This latter claim seems obviously true. If an eccentric super-genius were to gain complete control over my brain through no fault of 18 While the assumption that indeterministic free will is coherent is obviously objectionable, compatibilists are not helped by denying it; doing so simply ushers in the position that free will is impossible in any scenario. 16 my own, and if he were to force me to burglarize a house for him (against my will and despite my constant attempts to do otherwise), would I be morally responsible for either the burglary or the decision to commit it? Of course not. And A would likewise not be responsible for making the Bad Decision at t if P* had to override his brain and thereby make the decision occur. Therefore, since A has the alternate possibility of forcing P* to be the overriding cause of the Bad Decision, Mele and Robb fail to describe a scenario in which A lacks relevant alternate possibilities and is concurrently morally responsible; consequently, their P* example fails to disprove SPAP. Another new Frankfurt-style response is to imagine that C has managed to set up blockades in A’s brain that close off the final neural step for every decision other than the desired Bad Decision19 . C does this ahead of time, but this intervention needs not come into play since A could simply choose on his own to do as C wishes. I contest that this blockage response is inadequate since, as with the P* example above, A clearly has alternate possibilities. First, A could run into the neural blockades and force them to turn him away. If he does this, then his moral culpability seems to drop with each attempt to run a blockade. Second, and perhaps more importantly, A could completely refrain from making the Bad Decision by either remaining in a noncommittal deliberative loop or by continuously trying to run the blockades20 . The point to be drawn here is this: A would definitely have alternate possibilities available to him other than simply making the Bad Decision. Therefore, since this example must show that A is 20 Importantly, this ability to refrain from making the Bad Decision holds irregardless of the number and location of the blockades. Say there were so many blockades that every adjacent step was denied and the only neural move whatsoever that one could make would be down the path toward the Bad Decision. Even in such a scenario, it does not follow that one must then move down the open pathway. One could instead remain in a kind of neuronal suspension in which one is either intentionally not moving or is being constantly denied in attempts to take a first step down blocked pathways. 19 David Hunt, “Moral Responsibility and Unavoidable Action,” Philosophical Studies 97 (2000): 195-227 17 morally responsible and that he had no relevant alternate possibilities, the blockade example fails to disprove SPAP. Let us now turn to the question of whether it is even theoretically possible for a Frankfurt-style example to successfully disprove SPAP when A’s natural behavior is indeterministic. In general, if A’s natural behavior is indeterministic, then the only way for him not to have relevant alternate possibilities is via an intervention in which something else steps in and actively forces/overrides A’s decision. But if his decision is made for him via intervention in this manner, then he is clearly not morally responsible. Therefore, it is impossible for a Frankfurt-style example to successfully disprove SPAP if A’s natural behavior is indeterministic. At this point a frustrated compatibilist may want to assert that random indeterministic brain events are not worthy of being termed free decisions. But if this assertion is made, I must remind the compatibilist that, since moral responsibility implies free will, such a claim implies that people can’t be held morally responsible for their decisions. And if decisions don’t carry moral responsibility, then SPAP can’t be proven false and no progress can thereby be made toward showing that free will and determinism are compatible. I conclude this section by summarizing that if A’s natural behavior is indeterministic, then traditional, blockage, P*, and all other possible examples fail to present a case in which A is morally responsible and has no relevant alternate possibilities. Therefore, if A’s natural behavior is indeterministic and he is morally responsible, then the consequent of SPAP cannot be shown to be false and SPAP cannot be disproved. 18 Conclusion: As stated at the outset, in order to help compatibilism a Frankfurt-style argument must show that moral responsibility does not require alternate possibilities. And as we saw in Section II, in order to show that moral responsibility does not require alternate possibilities a Frankfurt-style argument must disprove SPAP. But we saw in Section III that no Frankfurt-style argument can disprove SPAP. Therefore, no Frankfurt-style example can show that moral responsibility does not require alternate possibilities. And therefore, no Frankfurt-style example can help compatibilism. So why have Frankfurt-style examples been so influential? I think that most people have an underlying assumption that their own behavior is indeterministic; that is, they assume that they have alternate possibilities in their lives and don’t believe that all of their behavior is entirely and unalterably determined by a particular physical state (including physical laws) of the universe long before their birth. I think that Frankfurt-style examples appear convincing on the surface to most people because most people project the view of their own indeterministic nature onto A and assume that his natural behavior is likewise indeterministic. That’s why they blame A for not forcing C to intervene: they think that A really could have forced C to intervene even if A’s universe has physical laws and an actual past extending back long before his birth. Folks with underlying determinist (but compatibilist) intuitions would also consider A morally responsible, albeit for a different reason. From their perspective, A is morally responsible because he acted rationally and free of any resolving intervention by a controlling agent. Therefore, it is entirely understandable that almost everyone21 would 21 One notable exception would be the folks who consider moral responsibility to be impossible. 19 feel an intuitive pull toward assigning moral responsibility to A. And this near-unanimous agreement about A’s moral responsibility is why the thought experiment has caused such a stir. But since the widespread agreement about A’s moral responsibility is generated due to differing assumptions among readers, this agreement does not help the compatibilist position. The apparent gains instead prove to be illusory, since no actual progress is made toward showing A to be morally responsible when not having any relevant alternate possibilities. And it is progress in this latter problem, not in moral agreement generated by ambiguous conditions, which will represent real progress for compatibilism. References Frankfurt, Harry, “Alternate Possibilities and Moral Responsibility,” Free Will anthology for the Oxford Readings in Philosophy, 2003, Edited by Gary Watson. Glymour, Bruce, and Marcelo Sabatés, “Micro-Level Indeterminism and Macro-Level Determinism,” Proceedings of the Twentieth World Congress of Philosophy, Volume X: Philosophy of Science, Edited by Tian Yu Cao. Hunt, David, “Moral Responsibility and Unavoidable Action,” Philosophical Studies 97 (2000): 195-227. Kane, Robert, A Contemporary Introduction to Free Will, Oxford University Press, 2005. Mele, Alfred, and David Robb, “Rescuing Frankfurt-style Cases,” Philosophical Review 107 (1998): 97-112.