You are on page 1of 19

Does Libertarian Freedom Require Alternate Possibilities? Author(s): Linda Zagzebski Source: Noûs, Vol.

34, Supplement: Philosophical Perspectives, 14, Action and Freedom (2000), pp. 231-248 Published by: Blackwell Publishing Stable URL: http://www.jstor.org/stable/2676130 . Accessed: 14/06/2011 12:47
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at . http://www.jstor.org/action/showPublisher?publisherCode=black. . Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Blackwell Publishing is collaborating with JSTOR to digitize, preserve and extend access to Noûs.

http://www.jstor.org

2000 DOES LIBERTARIAN FREEDOM REQUIRE ALTERNATE POSSIBILITIES? Linda Zagzebski University of Oklahoma According to causal determinismthe chain of causes of an act leads backwards in time to something completely outside the agent. we have identified who or what is responsible for it. Not every event in the . A is not determined. Of course. What bothers libertariansis that it seems to follow that something outside the agent is the ultimate cause of the act. the entire history of the world up to but not including the time T at which an act A occurs is compatible both with A's occurring and with A's not occurring. If the condition specified on the right side of TCP is met. what makes it happen. But whatever digression we take. One way to get criteriathat insure that an act is not causally determinedis to insist that there be no conditions existing just prior to the time of the act that are sufficient for the agent to do the act. Action and Freedom. then causal determinism seems to imply that no one is ultimately responsible for his acts. And if what ultimately makes an act happen is what ultimately is responsible for it. it is a major philosophical challenge to explain what is involved in making something happen and it is sometimes tempting to shift the discussion to something easier to grasp. 14. But TCP goes well beyond what is requiredfor A to be causally contingent. A could not be made to happenby anythingprior to it if both its happeningand its not happeningare compatible with everything priorto it. Call this the TemporalContingency Principle (TCP): (TCP) An act A at T is non-determined(temporallycontingent) if and only if there is a possible world W that has exactly the same history up to T as the actual world and in which A does not occur at T. we should not forget that once we have identified who or what makes an act happen. that condition is often taken to be constitutive of the denial that acts are determined.l That is. if anything is.Philosophical Perspectives. In fact.

Since the causal history of an event is a subset of its temporal history. we can only be sure that we have included everythingin its causal history if we include the entire past.we are ignorantof a great deal of the causal history of most events. for instance.232 / Linda Zagzebski temporal history of A is in its causal history. Therefore. he is not asking whether the rates would go down if everything happened in the past remainedthe same and that the money supply increased since most of what happenedin the past has nothing to do with interest rate increases. Isn't a causally contingent event a break in the causal order? Indeed. but the difficulty of the question should not lead us to conclude that the causal history of an event either includes the entire past or nothing at all. If an event satisfies TCP it also satisfies PCC. If the causal history of an event includes all events relevant to explaining what makes it happen. then all events have a causal history even if some events are such that their causal history is not sufficient to fully explain what makes them happen. as far as we know. we can therefore be sure that we have considered everything causally relevant to the occurrence of a given event only by relating it to the entire past.But perhapsit will be arguedthat TCP is not too strong since if any event A failed the condition of temporal contingency we would automaticallyjudge that it is causally deter- . Consider what we ask for when we want to know whether two events are related as cause to effect. TCP rules out the possibility that apparentlyirrelevantevents in the past causally necessitate an act A by requiring that nothing in the past is such that its occurrence rules out the non-occurrence of A. The strongerTCP is tempting because of our ignorance. Unfortunately. it is. but it is not necessary. anything in the past might be in its causal history.S. When someone inquires. since. The scope of the causal history of an event is a difficult question. of course. interest rate increases in 2000 are causally related to a distant military event in 1287. It is not an event that has no causally explanatorytie to the events precedingit. But TCP is tempting for anotherreason. it is doubtful that any event is such that the entire past history of the world is part of its causal history. "If the money supply increased. would the interest rates go down?". It is even possible that U. Perhapsit will be objected that if an event is causally contingent it has no causal history. but a causally contingent event is not a randomevent. TCP is a principle of safety even though it is strongerthan is required. TCP is too strong. The same point applies to the criterion for causal contingency. The principle we need instead is something roughly like the following Principle of Causal Contingency (PCC): (PCC) An act A is non-determined(causally contingent)if and only if there is a possible world W in which all the events in the causal history of A in the actual world occur and in which A does not occur.Temporalcontingency is sufficient for causal contingency. So even though the attempt to identify causes is an attempt to identify that part of the past that is its causal history. the invasion of Poland by the Tatarsin 1287. In fact.

it is causally contingent. Temporal contingency entails that the agent had alternate possibilities at the time of the act. the foreknowledge example illustrates the point. Accidental necessity is the alleged necessity of the past. Suppose that if Poland had not been invaded by the Tatarsin 1287 but everything else prior to a particularrise in interest rates remained the same. then when the agent does A she could have done otherwise. presumably. Causal necessity entails temporal necessity but not conversely. Again. Doesn't it follow that the invasion is in the causal history of the movement of interest rates in 2000? No. temporal contingency entails both causal contingency and alternate possibilities.3 In short. and if my act has no other determiningcauses. consider the fact that it is often thought that divine foreknowledge takes away human free will because of the accidental necessity of past infallible foreknowledge.Therefore. God's foreknowledge of my act might make it the case that I cannot do otherwise even though God's foreknowledge does not cause my act. if temporalnecessity is a problem for freedom and responsibility it is not because it entails causal necessity. A cannot but occur. and we are in no position to think that. if it turns out that a property requiring causal contingency does not requirealternatepossibilities.2If there is past infallible knowledge that A will occur in the future. But notice that this argumentdoes not imply that A is causally determined. there is no longer anything anybody can do about either the fact that the foreknowledge occurred or its infallibility. To take an historically importantexample of a temporalmodality that is not causal. it does not. .then. Causal contingency does not entail alternatepossibilities.and Possibilities/ 233 Libertarian Freedom Alternate mined. II One way to formulate the libertarianthesis on responsibility given in the first two sentences of this paper is the following: (LTR) An agent is morally responsible for her act only if the act is causally contingent. interest rates would not have risen.once an event occurs it acquires a form of necessity because there is no longer anything anybody can do about it. To think that it does follow is to think that all counterfactual relations between events imply a relation of temporal modality and that all temporal modalities are causal. It should not be surprising.Temporalcontingency entails causal contingencybut not conversely. Responsibility may be such a property. The thesis that the ability to do otherwise is a requirementfor responsibility is given in the following well-known principle: (PAP) An agent is morally responsible for her act only if she could have done otherwise. Causal contingency as defined by PCC does not entail alternatepossibilities. If it is compatible with the entire past history of the world that the agent does either A or not A. Hence.

In this section I will argue that F cases do not presuppose determinism. an evil neurosurgeon. Three decades ago HarryFrankfurtpresented a famous argumentthat was intendedto drive a wedge between responsibilityand alternatepossibilities. Black inserts a mechanism into Jones's brainthatenables Black to monitorand to controlJones's neurological activity. the mechanism will not intervene.In section III I will argue that F cases probably succeed in falsifying morally significant versions of PAP.But let us first review a standardFrankfurt-stylecase: Black.S The connection between responsibility and libertarianfreedom remains unbroken. Knowing that Mary Jones also despises White and will have a single good opportunityto kill him. ought not to be one of them. it cannot be because alternate possibilities are a requirementof causal contingency. then.7 Robert Kane and David Widerkerhave independentlyarguedthat F cases do presuppose a deterministic universe and hence cannot be used to support determinism. LTR may be true even if PAP is false.4In my judgment he succeeded admirablyin breaking apart the conceptual connection between responsibility and alternatepossibilities. Since causal contingency does not entail alternatepossibilities it is possible for an act to be causally contingent even when the agent lacks alternatepossibilities. If the activity in Jones's brain suggests that she is on the verge of deciding not to kill White when the opportunity arises.6 Frankfurtbelieves that his cases support determinism by falsifying PAP and that they do not presuppose a deterministic universe. Now suppose that when the occasion arises. she would have been thwartedby Black's device. Kane and Widerkeragree with Frankfurtthat the denial of PAP is closely connected with determinism. and in section V I will argue that F cases give no support to determinismand arguablysupport non-determinism. the kind of freedom that requirescausal contingency. On the other hand. I think that a careful look at Frankfurtcases actually supports non-determinism. and Frankfurtthought he could thereby drive anotherwedge between responsibility and libertarianfreedom.234 / Linda Zagzebski The reasons for holding PAP are no doubt different for different people. it appears that she is unable to do otherwise since if she had attempted to do so.On the contrary. Black's mechanism will intervene and cause Jones to decide to commit the murder. if Jones decides to murderWhite on her own. Noneand theless.Their disagreement is over the issue of whether . I have already mentioned that divine foreknowledge is one way in which that possibility might be actualized. but he went no distance at all towards supportingdeterminism. but I want to arguethat the desire to defend libertarianfreedom.Jones is morallyresponsiblefor her act.8 posiMy tion is that F cases do not presuppose determinismbut neither do they support it. In the judgment of Frankfurt most others. Jones decides to kill White without any "help"from Black's mechanism.wishes to see White dead but is unwilling to do the deed himself. If alternatepossibilities are required for responsibility.It will merely monitorbut will not affect her neurological function.

and Possibilities/ 235 Libertarian Freedom Alternate PAP can be falsified independentlyof falsifing determinism. it should be possible to falsify cases probPAPwithouteitherentailingor presupposingdeterminism. So the machine cannot tell for sure whether or not to intervene.the agent can make his decision prior to T2. and the other non-deterministicand resulting from conscious deliberationand choice. what happens if the agent makes the "wrong"choice at T2? Mele and Robb have thought of this problem and they stipulate that the machine overrides the agent's choice.9Their idea is to postulatetwo independentcausal processes leading up to the act. that no Frankfurt-style The problem as Kane and Widerkerpose it can be put briefly as follows: The Frankfurtmachine cannot operate unless it is possible to be a perfect predictor in a non-deterministicuniverse. The deterministicprocess and the indeterministicprocess coincide at the moment of decision. In their scenario the F-style machine does not simply monitor the agent's neurological activity. In this section I want to look at the Kane/Widerker argumentthat attempts to show case can falsify PAP without presupposingdeterminism. Since the process of deliberationis also indeterministic. Since there are two independentcausal sequences in operation in this scheme. but I would not claim it is impossible. But we are to imagine that he decides on his own to steal the car at T2. there are several possibilities to be considered. Since two contrary choices cannot occur simultaneously. Mele and Robb specify that in . More precisely. It is set to deterministicallycause the agent to decide to steal a car at T2. The most the machine can do is to make her change her mind. non-determinismdoes not entail PAP. Mele and Robb specify that the machine is so designed that in that case the deterministicprocess ceases operation and the indeterministic process is the one that is effective. Hence.Frankfurt ably succeed in doing that. but the deterministicchoice overrides the indeterminr istic one when they diverge. one deterministicand unconscious. some other agent-causedevent-whatever is the original undeterminedcause of her act. There are two other possibilities. One way out of the Kane/Widerkerobjection proposed recently is that of Alfred Mele and David Robb. then.the agent is already responsible. but even if they do not. the argumentof section I should lead us to expect that it is possible to construct cases that do.l° We need to accept. a Frankfurtmachine is impossible. It is not possible to be a perfect predictor in a non-deterministicuniverse. the event might or might not occur up until the moment it does occur. But the machine cannot intervene after the event since then it is too late to falsify PAP. consider whatever it is that the agent is responsible for according to the libertarian-an act of will. and that's not an F case. Suppose the agent decides on his own before T2 to steal the car. as I have argued. The second possibility is that the two processes diverge at T2. T2. the exact moment the deterministic mechanism is set to cause the same decision. Since. that the machine is designed in such a way that the indeterministicchoice overrides the deterministic one when they coincide. The machine cannot intervene before that event in a non-deterministicworld because no matterwhat the prior signs. This is a stretch. The first is the F-style scenario.

it is not clear exactly which processes/events Stump thinks are causally undeterminedand which are not. In Mele and Robb's scenario the agent can do otherwise at every moment up to but not including T2. She might easily have made it momentsearlier. She argues that if libertarianfree choices are correlatedwith temporally extended neural sequences.11 Suppose instead that the agent decides on his own before T2 not to steal the car. If the process of making the decision has already begun but the outcome is still undetermined.l3As Stewart Goetz has pointed out. the process of making a decision might itself be temporallyextended and temporallycoincide with a neurological sequence. The device is set to intervene whenever something in the agent tips off the machine that the "wrong"choice is ahead. it is not the case that the device insures that a deliberatorcannot choose otherwise without making her change her mind. It is only chance that the agent indeterministicallymakes her decision at T2.leading to a deterministicdecision to steal the car at T2 unless the same decision is made indeterministicallyat T2. but before a causally sufficient condition obtains. it is only chance that the scenario turns out to be like Frankfurt'sand it is only chance that the Mele/Robb scenario falsifies PAP.l4 But Stump's point seems to be that intervention does not require infallible prediction since all that is required is that it occur after a causally necessary condition for the choice (or other responsible event) has occurred. Alternatively. That is. the machinecan intervenewithoutneed- . For example. it is only chance that in the Mele/Robb scenario the agent is responsible for an act she performed at a moment when she could not do otherwise. These features of the Mele/Robb scenario make it importantly different from more standardF cases. The device is already in operation.236 / Linda Zagzebski that case the deterministicprocess in place causes him to decide to steal the car again at T2. an F-style machine can operate in a nondeterministicworld by interveningafter an initial neural sequence occurs which is a necessary condition for making the choice the machine does not want. the machine causes him to change his mind at T2 and to decide to steal the car after all. Since there is only a single moment at which the agent cannot do otherwise (assuming the description of what happens at T2 is possible). Mele and Robb say that if the device were to issue at T2 in the agent's deciding to steal the car. but it is more than an instant. That durationis generally assumed to be short.in which case the scenariowould not have been an F case.and if the decision cannot be completedwithoutsuch a beginning. and indeterminacy in this mattercan make her descriptionappearvulnerableto the Kane/Widerker objection. Hence.ls Either option should work. The machine can intervene after the neurological process begins but before it ends. Eleonore Stump has a different approachto answering the Kane/Widerker objection.then.the neurological sequence precipitating interventionmight be a causally necessary condition for the choice which occurs subsequently and is causally contingent. In the latter the agent cannot do otherwise during the entire period of time that the agent is deliberatingand making up her mind. it would erase any memories that are incompatible with its so issuing.l2Presumably.

and that is in principle possible in a nondeterministicworld.even if Black is an excellent judge of what Jones is about to do. All the machine need do is to intervenejust after such a moral reason occurs to the agent with the given force. The best description of a case of this kind I know of is given by Derk Pereboom. but it is considerably less than Widerkerand Kane suppose is required.Libertarian Freedom Alternate and Possibilities/ 237 ing to perfectly predict the outcome. he waitsuntilJones4 aboutto makeup So is his mindwhatto do. III Some libertariandefendersof PAPcontend that the agent in Frankfurt cases does have ethically importantalternatepossibilities. I conclude that the description of standardF cases does not presuppose determinism. All that F cases require is the capacity to identify a causally necessary condition for a non-determined choice in advance of the choice. even though she does not have the alternative of .l8In Pereboom's case the machine is set to make the agent decide to declare an illegal tax deduction if he does not do so on his own. The agent's psychology is such that a causally necessary condition for him to decide not to commit the illegal act is that a moral reason occur to him with a certain force. this means the machine might intervene unnecessarily.The fact that most commentatorson F cases have posed them in terms of an infallibly predictive device has no doubt misdirectedsubsequentdiscussion. This is a bit more than Frankfurthas a right to claim about the case. But the case requiresthat he be infallible only with respect to identifying a causally necessary condition. he is not infallible. For the same reason.but he prefersto avoidshowinghis handunnecessarily. Frankfurt'sown description of his principal In case suggests this interpretation well: as Supposesomeone-Black. Blackis prepared go to considerable to lengthsto get his way.Such a view not only does not presuppose determinism. but that is not a problem. Of course. alternatepossibilities entailed by her responsibility.l7 Presumably. and he does nothingunlessit is clear to him (Black is an excellent judgeof suchthings)thatJones4 goingto decideto do something is other thanwhathe wantshim to do. Stump's description of F cases in an earlier paper suggests this interpretation. The machine can thereby make it the case that the agent cannot do otherwise even in a non-determineduniverse. the machine can intervene after any necessary condition for making the "wrong"choice has occurred. let us say wantsJones4to performa certainaction. He can infallibly identify neither causally necessary nor causally sufficient conditions for Jones' choice.'6 fact. What makes Pereboom's case especially promising is that many libertarianswill agree that thinking of a moral reason is a causally necessary condition for choosing the morally right thing in many circumstances.but it is likely to be part of a typical libertarianaccount of what usually happens in cases of moral temptation.

a number of variations of PAP have been proposed and defended against Frankfurt-styleexamples. Whatever is the effect of active power must be something that is contingent.According to this view the cause of an act for which the agent is responsible is the agent herself. then. but because the power of agency itself demands it. it may be sufficient to take away the principalmotive of some libertarians for finding a true version of PAP. Thomas Reid contended that it is essentially a two-way power. they then feel compelled to look aroundfor some variationof PAP that is true and that can withstand Frankfurt'sattack. not because causal contingency entails alternatepossibilities. Libertarians are not committed to a single position on the process whereby an agent causes her act. We can conceive no way in which power may be determinedto one of these ratherthan the other. in a being that has no will. what makes it happen. So for them I need not say anything more. but there is agreementthat the chain of causes of a free act ends in the agent.l9 Here Reid ties the possession of alternatepossibilities to agent causation itself and he seems to be inferring the contingency of an act from the fact that it arises in a situation in which the agent had the power not to perform the act. Otherlibertariansmay accept my position that PAPis not entailed by causal contingency and yet continue to maintain that PAP is entailed by libertarian freedom since there is more in libertarian freedom than causal contingency. not an event of any kind. Since they are convinced by F cases that PAP is false. Contingentexistence is that which dependedupon the power and will of its cause. is an argumentconnecting a form of the libertarianposition on freedom with alternatepossibilities. This position requires an argument going from a substantive account of the nature of libertarianfreedom to the consequence that the agent has alternate possibilities. An agent does not have the power to cause an act unless she also has the power not to: Power to produce any effect implies power not to produce it. One way in which this position has been explicated is via the concept of agent causation. they will be forced to give up belief in libertarianfreedom. The argumentis that agent causation entails alternate possibilities. Let us call this Reid's principle: . It is obvious that some libertariansdefend PAP only because they believe that libertarianfreedom requires alternatepossibilities. not even the event of her exercising her agency. From this point of view agency is a kind of power. they think. Consequently.238 / Linda Zagzebski doing otherwise. Otherwise. Here. But if the argumentI have given that causal contingency does not entail alternatepossibilities is convincing. whereas PAP is a thesis about what the agent might have caused instead. There is nothing outside the agent that is ultimately the cause of her act.20So on Reid's view agent causation is partially constituted by the possession of alternatepossibilities. What would such an argumentlook like? Determinismand its denial are theses aboutthe causes of an act.

She can't do anything to avoid being responsible for the act. RP can be used to defend a form of PAP. But it seems to me that even though it was open that she not make A happen. The machine does it for her. Wierenga'sversion is the following: (PAPC)Necessarily. S for is morally responsible performing only if S couldhaverefrained for A fromcausing A. It is true that either the act is agent-caused or an act of the same type is machine-caused.23McKenna does not specifically mention agent causation either. the machine intervened before she could exercise that power. everypersonS andact of will A suchthatS causesA. If S makes A happen. Of course in that case S would have exercised the power not to do A if the machine hadn't intervened. but his discussion of the power of an agent to make the act her own or not is clearly similar in spirit to the views of Reid and Wierenga.Libertarian Freedom Alternate and Possibilities/ 239 (RP) An act A is agent-causedby S only if S had the power not to agentcause A. who argues that in F cases it is up to the agent whether or not her act puts her stamp upon the world. It was open to the machine to make her make A happen.The argumentwould go as follows: (1) An agent S is morally responsible for an act A only if S agent-causes A.24But even then it . The most she can do is to fail to will. it was not open to her not to make A happen. (3) An agent S is morally responsible for an act A only if S had the power not to agent-cause A. that "could have refrained"in PAPC does not refer to a power that S has. but even then it is the machine that makes it happen that she fails to will. But as it turned out. Wierenga claims that even in F cases S could have refrained from causing A.then it was open to her to make it happen or not to make it happen. then. (RP) Therefore.21 It is clear from Wierenga's discussion that he has agent causation in mind.It is also true that in most descriptionsof F cases the machine's operation depends upon what the agent does. (3) is a form of PAP. It appears. (2) S agent-causes A only if S had the power not to agent-cause A.22 A similar defense of alternatepossibilities has been given more recently by Michael McKenna. A version of PAP very similar to (3) has been defended by EdwardWierenga as a response to F cases.

McKenna makes another move in an attemptto preserve the idea that the agent in F cases retains a two-way power. But although I think that doubt is in order. it is before she can exercise an alternatepower. The power to open my eyes is the kind of power I need to be able to exercise now if I possess it now. makingbread. in the case of making bread. if my eyes are now bandagedshut. He says an agent can have a power even though she is unable to exercise it.It is only because of what she wouldhavedonethat the machine goes into operation. In any case. The fact that I cannot initiate the process at this moment is of no consequence. the machine operates before she is in a position to exercise any contrarypower. Might it still be the case that she has the power to choose otherwise even though the machine prevents her from exercising it?25 I am inclined to think we should be as suspicious of unexercisable powers as of imperceptibletrees. or writing one's will is not compromisedby the fact that a piano. The fact that I cannot exercise the power right now does not detract in any significant way from my present possession of the power since it is not the kind of power that requirescontinuousexercisability. My conjecture is that the possession of a power at T and the ability to exercise it at T split apart only for some acts that are clearly not basic acts. I do not now have the power to open my eyes. The power to performsuch complex activities as playing a SchubertImpromptu. the time it takes is discontinuous. what is gained is not very helpful to the defender of (3) or PAPC. In Pereboom's example the machine does operatebecause of what the agent actually does-thinking of a certain moral reason. even if I am wrong in this conjectureand an agent can possess the power to refrain from choosing in F cases without having the ability to exercise that power. The power to refrain from choosing an act is like the power to open my eyes and unlike the power to make bread. but the power need not be exercisable? An unexercisable power is a pretty thin reed upon which to hang moral . Choosing and refraining from choosing are basic acts and if my conjecture is right then. whereas playing an Impromptuand making bread are far from basic acts.I actually agree with McKenna that there are powers one can have without being able to exercise them at every moment at which the power is possessed. That is probably because opening my eyes is a basic act or very close to a basic act.240 / Linda Zagzebski is not up to the agent whether or not the machine operates.These activities take some time to performand. like opening my eyes. What is the point of insisting that an agent is responsible for her act A only if she has the power to agent-cause A. kitchen equipment.If the machine operates. Again. they are possessed at T only when I can exercise them at T. or writing materials is not available at this moment. In contrast. But having a moral thought is not exercising the right kind of power for attributingmoral responsibilityif indeed it is exercising a power at all. I have said that in F cases the machine takes away the agent's ability to exercise the power to choose otherwise. One does not at T have the power to perform a basic act unless one also possesses at T the ability to exercise that power.

although it is open that she not make it happen.Possibilities/ 241 and Freedom Alternate Libertarian responsibility. more enduringpropertyof the agent that is possessed even in some situations in which she cannot do otherwise. McKennasays he's willing to concede that the agent does not have it within her power to do otherwise. Now if the machine intervenes after a necessary condition for the . She fails to exercise agent causation. It is not avoidable throughthe exercise of her own power. not (3). S cannot exercise a power to refrain because she can never complete the act of deciding to make what the designer of the machineconsidersthe wrong choice. But the avoidability of S's agent-causing A in (3') amounts to nothing more than that it might not happen. (3') does affirm an alternatepossibility of a weak kind. Another approach to connecting alternate possibilities and responsibility focuses on the conditions under which an agent is blameworthy. and that does not add anything to the thesis that it is causally contingent. the designer of the machine is the cause of S's not causing A. althoughit is avoidablebecause of something about her or something she does. the strongest principle we can affirm is not (2) but (2'): (2') S agent-causes A only if S's agent-causingA is avoidable.27The advantageof Otsuka'sproposal is that he does not make it a requirement for blameworthinessfor performing an act of a given type that the agent could have refrainedfrom performingan act of the type for which she is blameBut she must have had an option open to her that gives her a way out worthy.28 of blame. the state of affairs of her agent-causingher act is avoidable. If S refrainsfrom causing A.It is not open to her to make it not happen. Since the most we have a right to say in F cases is that her agent-causingA is avoidable.26Now my remarks above imply that there is a sense in which this is true.What McKenna may be trying to point to is a deeper. And if she cannot exercise the power to refrain she does not have the power to refrain. but failure to exercise agent causation is not the same thing as exercising agent-refraining. And we can conclude. he says she could have avoided the particularthing that she did. I have arguedthat her status as a libertarianagent can remain even in such cases and is that deeper property. Nevertheless. although it is due to a power she would have had and exercised if the machine had not intervened. and since she does agent-cause A in F cases. but (3'): (3') An agent S is morally responsible for an act A only if S's agentcausing A is avoidable. I conclude that PAPC is false because it is false that the agentcould have refrainedfrom causing A. Michael Otsuka argues that an agent is blameworthy only if there is something she could have voluntarily done instead that would have rendered her entirely blameless. Even though she can't avoid agent-causing her act. The avoidability is not due to a power that S has.

but its avoidability is not due to her own power.Nonetheless. and yet we think she is to blame if the machine does not need to intervene. F cases can be interpretedas counterexamplesto (2) as well as to most versions of PAP. thereby renderingher blameless. contra Kane and Widerker. (2') can be supportedand used to conclude (3'). not because she might have had a moral thought which would have been necessary but not sufficient for deciding not to commit the act. If the agent could have entertainedthe thought voluntarily. However. not that the agent must have been able to do something to avoid it.I have also arguedthat it is likely that successful versions of F cases can be devised that do not require that the machine be a perfect predictorof the agent's choices. the challenge facing defenders of any version of PAP is that the version defended not only must not be falsified by F examples.The beauty of these thought experiments is that they force us to confront what it is in a situation in virtue of which we judge the agent responsible. when the machine does not need to intervene and in ordinarysituations in which there is no machine in existence we do not blame her because she could have had such a thought. but it has to be motivated. In short. Even though there almost always are alternate powers of some sort in cases of agent causation. he can affirm (3'). it is doubtful that they are an intrinsic part of the power of agency itself. is that we . even though I think it is right to focus on agent causation in identifying the ground of responsibility.Otsuka is right that her blameworthiness is avoidable. but her blamelessness would have little to do with what she did.On the other hand. IV I have claimed it is likely that no significant version of PAP can be devised that is not falsified by Frankfurt-stylecounterexamples. She is to blame because her act was the result of a libertarianfree choice. Having a moral thought is not exercising the kind of power relevent to her blameworthiness. I believe. Hence. my real reaction to F cases does not depend upon their being literal counterexamplesto PAP.Suppose that we use a case like Pereboom'sin which the precipitatingevent is a moral thought which is necessary but not sufficient for the agent to make the decision the machine wants to prevent. Like the other defenders of alternatepossibilities. Surely.242 / Linda Zagzebski choice the machine is designed to prevent. If the thought is a kind that always occurs involuntarily. but he cannot attributeany significant form of two-way power to the agent. but that tells us no more than that a responsible act must be avoidable.then it is true that there is something she could have done that would have precipitatedthe action of the machine. The problem is that the motivation is likely to go through an account of agency that utilizes something like (2). What we see in an F situation. it is probably wrong to think of the power of agent causation as involving alternate powers.there is nothing the agent could have done before the intervention that would have renderedher blameless. the interventionis too early to satisfy Otsuka'srequirement.

a way that calls our attention to the difference between the features of a responsibility situation that are salient and those that are not. There is no doubt that the mechanism has some significant power of constrainton the agent. McKenna. The first and weaker point is simply that the libertarianis not forced to give up LTR when she accepts that the agent is responsible in F cases. We do not evaluate Jones any differently with the machine than without it. There are some alternatepossibilities which she normally possesses in a non-deterministicworld which she does not possess on the machine.The second and stronger point is that she will accept that the agent is responsible in F cases for libertarian reasons. so is the latter. nor do we attributeresponsibility to her in any different way once we think we have identified the precise sense in which she has alternatepossibilities. and if the former act is causally contingent. Since the point of the F cases is to show us the irrelevance of the mechanism. Either way she is responsible.then. in fact. it does not matterhow far the power of the mechanism actually goes.they do not give any support to determinism. have alternatepossibilities of some kind in Frankfurtsituations. Suppose that Wierenga. because doing so has no effect on the deeper libertar- .The avoidability of the one is the same as the other. Notice that this point is not the same as the first. nor would we evaluate Jones any differently if. Libertarianism is compatible with the rejection of PAP. What the F cases do is to make us think about responsibility in a novel way.LibertarianFreedom and Alternate Possibilities / 243 don'tcare what Black's mechanism is capable of doing because it doesn't actually do anything at all. we see that we don't care whether or not the agent has alternatepossibilities. V The intuition that the agent is responsible in F cases does not refute the libertarianthesis LTR. She accepts F cases because the libertariancondition still obtains in those cases. She can reject PAP.This means that F cases do not refute the libertarianposition. and others are right in thinking that she does. And since the lack of alternatepossibilities is tied to what the mechanism is capable of doing ratherthan to what it actually does. nor does it falsify (3'). But the intuition of most people who hear of F cases is that the machine has no effect at all on her level of responsibility or on that for which she is responsible. What we see is that Black's machine plays no role in our judgment about Jones's responsibility. the machine became a perfect predictor. Furthermore. Otsuka. Whatever reason the libertarian has for thinking a person agent-causes his act in an ordinary non-Frankfurtstyle situation applies just as well to a Frankfurtsituation.That is because a certain kind of libertarianwill say that the reason why the agent is responsible in F cases is that she agent-causes the act and agent-causationrequiresthat the act be causally contingent.Nor need we work to identify the respect in which the agent retains an alternativesince our judgment of her responsibility does not depend upon the success of that search. per impossibile (perhaps).

so in one clear sense of "because. This means that the comparison of the F situation with one in which the agent can do otherwise is importantto the intuition that the agent is responsible on Frankfurt'sown account. Frankfurtremarksthat it is importantthat there is no morally relevant difference between the situation in which the agent has alternate possibilities and one in which they have been eliminated by a Frankfurt-type mechanism. But notice that this implies that what the agent would have done if he had been able to do otherwise is relevant to his responsibility in the actual situation since the subtractionof the conditions that make it the case that he cannot do otherwise turns the situation into one in which he can do otherwise. so the comparison of the actual situation with one in which he can do otherwise is a comparisonof the actual world with a very distant world. it is a different and weaker position. wouldhave led him to do it or madehimdo it even if it hadbeenpossiblefor himto do something instead.. Now if someonehadno alternative performing certainactionbut did not do it becausehe was unable to a to do otherwise.244 / Linda Zagzebski ian intuitionoperativein these cases. he does not satisfy the condition for responsibility in the passage just quoted. Of course. On the page following the passage just quoted Frankfurtoffers an amendment to his claim and a new principle of alternatepossibilities which he can support:"a person is not morally responsible for what he has done if he did it only becausehe could not have done otherwise.She can capitalize on F cases to support non-determinism. thenhe wouldhaveperformed exactlythe sameactioneven if he could havedoneotherwise. A curious feature of Frankfurt'spoint here is that as a determinist he cannot claim that a possible world in which the agent can do otherwise is close to the actual world. or thatmadehim do it.29 else In a deterministic world the agent's act is caused by events outside the agent.30Notice thatthis is not simply a clarificationor innocuous additionto the passage quoted above."the act occurs because of those events. Whatever was thatactuallyled the or in it personto do whathe did. He says that the circumstancesthat make it the case that the agent could not have done otherwise could have been subtractedfrom the situation without affecting what happened or why it happenedin any way. The circumstances madeit impossible him to that for do otherwise couldhavebeensubtracted fromthe situation without affectingwhat happened why it happened any way. Hence.it notbe the case thathe the may actedas he didbecause he couldnot havedoneotherwise.but that is just because the deterministand the non-deterministsee every case differently. one of Frankfurt'sown remarksprovides the argument. To say that an agent is not responsible for an act if he did it only because he could not have done otherwise ." [emphasis added]. the deterministwill see F cases differently.. one in which the basic structureof naturallaw differs from that which actually obtains. Frankfurtsays: Eventhough personwas unableto do otherwise.The point is that F cases do not give the libertarianany more reason to accept determinismthan ordinarycases do.Ironically. But the libertariancan go farther. He acts because he could not do otherwise.

And if he had had alternatepossibilities he would have done the very same thing in the same way. in a deterministicworld it is not the agent that ultimately cases do nothing to lead us to rescind the view makes her act happen.What they do show is thatPAPis a false path. and that is the idea of causing an act. of naturein such a world. In Frankfurt cases in a non-deterministicworld the agent makes her act happenin as ultimate a sense as you like. We think that Frankfurtwas right in his point that the agent is responsible because what he does in an F situation is relevantly similar to what he would have done if he been able to do otherwise. Frankfurt that this is a significant difference.. Once we have identified who or what makes an act happen we have identified the potential bearerof responsibility for it. As a determinist. therefore.and Possibilities/ 245 Libertarian Freedom Alternate leaves open the possibility that when he is responsible he did the act partly because he could not have done otherwise.Frankfurtmust countenance dependence of the agent's act on the conditions that make it the case that he cannot do otherwise since on his view those conditions are the cause of the act. He is. The presence of alternatepossibilities may be a reliable sign of the presence of the agency needed for responsibility. making it happen. and if he had not existed the agent would have had alternatepossibilities. we get the conclusion Frankfurtwants precisely because of our interest in what would have happened if Jones had been able to do otherwise. It can help illuminatethe idea of of causing or making something happen. At the beginning of this paper I remarkedthat it is tempting to bypass the difficult concept of what it is to make an act happen and to take various digressions. There is nothing wrong with digressions as long as they illuminateour target concept and we eventually get back to it. He can't get off the moral hook thateasily. In that case. . Therefore.just as responsible as he would have been if he had had alternate possibilities.however. it presumablywouldaffect what happenedor why it happened in some way. but eventually we have to get back to the idea those conditions aimed to elucidate. But it is also importantthat Black's meddling is a quirk that in the normal course of events would not have entered the picture at all. Talk of counterfactualconditions for the attribution responsibilityis a digression. To say otherwise is to permit the agent too great a degree of positive moral luck. if the circumstances that make it the case that the agent could not have done otherwise were subtractedfrom the situation. Libertarianswill make a differentjudgment. Jones would have actually been able to do otherwise. But it is worth noting that this position requires a partial retractionof his claim that in a responsibility situation the conditions that make it the case that the agent cannot do otherwise are irrelevantto why he did what he did. The point could be put as follows: It is only an accident that Black exists..but it is not necessary for it. along with our inclination to think that but for the short-lived existence of an inoperativedevice. the second condition is not satisfied and the first is logically problematicsince any proposition of the form If Jones then. contraryto what Frankfurtsays in the quoted passage. If the universe is causally determined.has antecedentthat contradictsthe laws an hadbeenable to do otherwise.

ited by John MartinFischer (Ithaca. but it is not essential to it.246 / Linda Zagzebski I suspect that the moral of Frankfurtcases can be generalized to apply to many properties that have counterfactualconditions for their application. and Frankfurt'sAttack on the Principle of AlterDavid Widerker.For example. "Alternate furt does not conclude the truthof determinismin this paper. Possibilities and Moral Responsibility. it can happen that they fail because of the presence of a counterfactual manipulatoreven when the target property still obtains.Y. pp. pp.Y.3'In general. HarryFrankfurt.: Cornell University Press. . The necessity is said to be "accidental" because the event it is not necessary at all times but only after the event trans. If I am right that such conditions are usually proposed because they are a sign of what we think is really important. provided that it is the kind of event for which we attributeresponsibility to the agent in the basic sense. An early use of this type of 89 example appearsin "Responsibilityand Control."Libertarianism Review104 (April 1995). For a recent proponent of this position see David Hunt. The term "accidentallynecessary"comes from Ockham's use of the idea of necessity per accidens. an overt act. 247-261." forthcoming in Philosophical Possibilities and Moral Responsibility. pp. 6. several well-known definitionsof knowledgeinclude counterfactual conditions. Robert Kane. 1997). 1969). and Unavoidable Action. 1996). and in "Foreknowledgeand Freedom. 119." Philosophical ertarian Freedom and the Avoidability of Decisions. chap. there is nothing blocking the acceptance of determinism."Alternate edlosophy66 (December 1969). I have arguedthis before in TheDilemma Freedom Foreknowledge. 828-839. TheSignificance Free Will (N. p. I use "act"for the event for which an agent is primarilyresponsible. Way Out. note 3.but he thinks that PAP is the only thing standing in the way of the acceptance of the compatibility of determinism and responsibility. 149. Frank7. and God'sForeknowledge. It may be an act of will. In anotherplace I have argued that epistemic Frankfurtstyle cases show that these conditions can fail for reasonsparallelto the reasons for the failure of PAP."Journalof Philosophy (January 1982)."Journalof Phi4.a form of temporal necessity.: Oxford University Press. and "Libnate Possibilities. Subsequentpage references will be taken from the Fischer reprint."Companion Philosophyof Religion.causal determinismand non-determinismare theses about what did happen. FutureContingents. 1991). what might have happenedis often a good indication of what did happen. 1986). 24-40."FaithandPhilosophy #1 (January1999). It can then be argued that once compatibilism is accepted. of 8. and of 5.edited by Philip L. by Marilyn McCordAdams and Norman Kretzmann(Indianapolis:Hackett. See Predestination. "On Augustine's 16. occurs. p." Faith and Philosophy12 (January1995). N. (Oxford Univerand of 3."in Fischer.32PAP is a thesis about what might have happened. Quinn and Charles Taliaferro(Blackwell. pp. Notes 1. and "MoralResponsibility Studies. I made this point in TheDilemma Freedom Foreknowledge to sity Press. 143152. This adaptationof Frankfurt'sexample using a neurological device is similar to some of the cases described by John Martin Fischer. 6. 2. or even a firing of neurons. reprintedin MoralResponsibility.

p. Mele and Robb. 23. Whatis especially interestingaboutthis view is thatit makes agency and the possession of alternate possibilities the prior idea. See "Of the Liberty of Moral Agents. "Of the Words Cause and Effect. "Libertarian Freedom and the Principle of AlternatePossibilities."Philosophical Review 107. "AlternatePossibilities and Moral Responsibility. She argues that all it takes to falsify PAP is a device that interrupts the neural sequence correlated with an act of will after it begins but before it ends. 73-88. 14. 85. TheNatureof God:An Inquiry into DivineAttributes (Ithaca: Cornell University Press. pp. Eleonore Stump. "Of Active Power in General2" Essay 1 of Essayson theActivePowerof the Mind (MIT Press. vol." Faith and Philosophy16. "Intellect. 75. and Frankfurt: A Reply to Goetz.Libertarian Freedom Alternate and Possibilities/ 247 9. This interpretationof Stump is confirmed in "Dust. pp." Philosophical Studies46 (1984). 15. Reid thought that the idea of a cause derives from that of a power and concluded that in the strict sense only conscious beings are causes." in Faith. Determinism. of vol. 11. 85. This scenario falsifies PAP. In this paper Stump makes her position clearer. and it is even possible that they are identical. pp. Wierenga says his view is similar to that of Reid in the footnote on p. pp. 13. In fact. 1969). p.but the point is sound since we can easily imagine how someone could identify a causally necessary condition in a non-deterministicuniverse. Stump maintainsthatAquinas also rejectedboth PAP and causal determinism. ."pp. This implies that the intervention would occur after a causally necessary but insufficierit condition for deciding not to commit the murderhas occurred. and of edited by Michael D. 1989). 12. 28. P. 22. Causation and contingency are derivative. 10. #1 (January 1999). #3 (Winter 1997). 1996)." chap." Faith and Philosophy. The precise nature of the relationship between the act of will and the agent's brain states does not matterfor the argument. 148-149. 20. Of course. Beaty (University of Notre Dame Press. 16. 18. "Stumping for Widerker. 97-112. #3 (July 1999). 1969).(MIT Press. "AlternatePossibilities and Causal Histories. See Eleonore Stump. even if it is identical with a neurologicalevent. footnote 11. Will. edited by Jeff Jordan and Daniel HowardSnyder (Lanham. "RescuingFrankfurt-Style Cases. 2. Action and Active Power." in Essayson the ActivePowersof the Human Mind. Michael McKenna. but it does not require that an act of will is determined. and Rationality. 83-89. pp. Edward Wierenga. Freedom. 35. 255)."Journal SocialPhilosophy.Md: Rowman and Littlefield. In this paper Stump describes an episode from Dostoevsky's ThePossessedin which Verkhovenskyis preparedto intervene with a surgical procedure to insure that Fedya murdersthe Lebyatkins if there is "any chance" that he won't do so when offered a bribe (p. 21. Verkhovenskyis not in a position in identify causally necessary conditions with 100% accuracy. 19. "AlternatePossibilities and the Failure of the Counterexample Strategy. #1 (January1998). 267ff. p. An earlier version of the same type of response is given by MargeryBedford Naylor in "Frankfurt the Principle of Alteron nate Possibilities." this volume. 16. 249-258. 17." in Christian Theism the Problems Philosophy. 103. 1990).vol. 413-422. ibid. and the Principle of Alternate Possibilities. pp.

685-701. Otsuka makes this explicit on p. 30. Frankfurt. forthcoming in Virtue Epistemology. Oxford University Press. McKenna. and Daniel Speak for comments on an earlier draft of this paper. pp. p. 27. p. Frankfurt. 26. Calif. presented at conference on Virtue and Duty in Epistemology. 152. 28. I am gratefulto David Hunt." Ethics 108 (July 1998). Nov. McKenna.248 / Linda Zagzebski 24. 25. 29. 31. 690. 81. 32.pp. An exception is causal over-determinationscenarios in which the machine is set to cause a certain choice but the agent causes the same choice by her own power at the same time. 150-151."Incompatibilismand the Avoidability of Blame. Michael Otsuka. In "Must Knowers Be Agents?". . 77. 1999.Andrew Eshleman.p. 13. Santa Barbara.