1. The Fragilist Account of Alternative Possibilities
Harry Frankfurt famously argued against what is known as the Principle of Alternate Possibilities (or PAP): An agent is morally responsible for his action only if he could have done otherwise than perform that action.Footnote 1 Here is a much-discussed example that accords with Frankfurt’s line of reasoning:
Murder
Unbeknownst to Jones, Black has implanted a microchip in Jones’s brain whereby Black can completely monitor and control Jones’s inclinations toward forming his intentions and the resulting behaviours. Black does not always exercise the power that he has over Jones. However, Black can exercise that power whenever he chooses to do so. Under these circumstances, suppose that Jones intends to kill Smith. In furtherance of this intention, Jones aims at Smith’s heart, shoots him, and thereby kills him. As it happens, Black also wanted Smith dead. Thus, Black did not exercise his power when Jones decided to kill Smith. However, if Jones was going to decide at the last minute not to kill Smith, Black would have intervened and used his power over Jones to make him shoot Smith.
Given the aforementioned description of the story, it may seem that Jones could not have done otherwise but kill Smith.Footnote 2 Nevertheless, Jones does seem to be responsible for killing Smith because Black’s intervention that would have caused Jones to decide to kill Smith were he not going to initially form the malignant intention did not actually take place. Jones would have killed Smith anyway, even if Black had decided not to get involved in this homicide at all.
A number of incompatibilists have responded to Frankfurt’s counterargument by claiming that the action (or the event immediately occurring as a result of that action) performed by the agent in the actual scenario is not identical to the corresponding action (or the event that would immediately have occurred as a result of that action) in the counterfactual scenario. For instance, Peter van Inwagen argues that in the story of Murder, Jones in fact has alternative possibilities in the sense that he could have prevented the event that actually happened.Footnote 3 In the actual scenario, Jones performed an action (shooting Smith) and thereby caused an event: Smith’s death. Let ‘e’ stand for this particular event-token. Also, let ‘e´’ stand for the event-token that represents Smith’s death in the alternative scenario where Jones would have been made to shoot Smith via Black’s intervention. On this picture, it is sensible to say that e flows causally from Jones’s freely firing the gun via his own intention, whereas e´ is caused by Jones’s firing the gun via Black’s manipulation of the implanted microchip. Drawing upon the Davidsonian account of event-individuation (which holds that causes of an event are essential to the individuation of that event), van Inwagen argues that e is numerically distinct from e´, though these two events, taken in themselves, have intrinsic features that look exactly similar. Thus, it was in Jones’s power to prevent e from occurring by having tried to decide not to shoot Smith. Of course, in that case he would have ended up generating e´, which may seem exactly similar to e. Yet, as long as e´ is distinct from e, it is fair to say that Smith could have prevented e.
Similarly to van Inwagen, David Widerker says that Jones “can avoid the performance of his actual act of killing Smith, or can bring about the non-occurrence of that act” because the intention and the resulting act Jones committed under the influence of Black in the counterfactual scenario “would be a different intention from the one he actually formed, and a different act from the one he actually performed.”Footnote 4 More recently, Helen Steward argues that “Jones could … have done other than perform the very shooting he in fact performs” because “no possible world in which it is Black, rather than Jones, who initiates Jones’s φ-ing, is a possible world in which the same individual action occurs as [it] occurs in the actual world.”Footnote 5 These responses commonly advert to the non-identity of the actual and the corresponding counterfactual actions (or of the two events immediately caused by each of them) in order to leave some alternative possibilities open to the agent.
This strategy of adverting to the non-identity of the two actions (or events) in question can be understood in light of David Lewis’s discussion of fragility in the individuation of events. Lewis says that an event is fragile “if, or to the extent that, it could not have occurred at a different time, or in a different manner” and that a fragile event “has stringent conditions of occurrence.”Footnote 6 In terms of this characterization, the aforementioned incompatibilists can be said to adopt a fairly strong standard of fragility in individuating an action (or an event that is caused by an action), which enables them to differentiate the actual action (or the resulting event) from the appropriate corresponding action (or the resulting event) in the relevant counterfactual scenario. For this reason, I shall call the preceding incompatibilist strategy the fragilist account of alternative possibilities.
This paper deals with one particular argument against the fragilist account developed by John Martin Fischer. Fischer claims that the notion of freedom that figures in the fragilist defense of PAP (or the way fragilists employ the notion of freedom that figures in their defense) does not represent the appropriate sense of freedom involving alternative possibilities.Footnote 7 He considers a new formulation of the principle that he thinks employs the notion of freedom appropriately grounding moral responsibility, and argues that fragilists must accept this new formula. However, argues Fischer, doing so would not save fragilists because this formula is vulnerable to typical Frankfurt-style counterexamples such as Murder. As a result, Fischer claims, the fragilist project of associating moral responsibility with freedom to do otherwise is doomed to fail. In this paper, I shall come to the aid of the fragilist, and maintain that Fischer’s counterargument is unsuccessful. First, I shall argue that, when closely analyzed, the notion of freedom that figures in the formula Fischer provides is not undermined by the story of Murder. Then, for Fischer’s sake, I will present a more resilient kind of counterexample that does undermine the new formula. Finally, I propose a promising revision of Fischer’s formula that is resistant to the sturdier counterexample.
2. Fischer’s Critique of the Fragilist Account
The core of the fragilist account is to secure the agent some leeway for alternative possibilities by drawing a distinct line between his action or the resulting event in the actual scenario and the corresponding action or the resulting event in the relevant counterfactual scenario. On this account, Jones in the story of Murder has alternative possibilities because Smith’s death that would have occurred in the counterfactual scenario is a numerically distinct event from Smith’s death that actually occurs (or, in terms of actions rather than resulting events, Jones’s pulling the trigger in the counterfactual scenario is numerically distinct from his pulling the trigger in the actual scenario). Jones is free in the sense that there is an alternative future open to him where he would have avoided bringing about what he actually brings about (or where he would have refrained from doing what he actually does).
Fischer thinks that this fragilist picture is mistaken. He holds that on the proper understanding of the freedom involving alternative possibilities, the agent must act freely in each of the relevant alternative possibilities.Footnote 8 He writes:
On the traditional alternative-possibilities picture, it is envisaged that an agent has a choice between two (or more) scenarios of a certain sort. In one scenario, he deliberates and forms an intention to perform an act of a certain kind and then carries out this intention in an appropriate way. In at least one other possible scenario, he deliberates and forms an intention to perform a different kind of act (or no act at all) and carries out this intention in an appropriate way. This is what is involved in having robust alternative possibilities, and certainly this is the natural way to think about the sort of alternatives that allegedly ground moral responsibility.Footnote 9
Fischer notes that the fragilist defense of PAP fails to employ what he takes to be the traditional, robust sense of freedom involving alternative possibilities. According to him, when we say that an agent was free to perform an action (in the sense involving alternative possibilities), we normally require that there be an alternative action that he could have performed “as a result of his character or practical reasoning” other than the one he in fact performed.Footnote 10 In other words, “for the agent to have deliberate control (in the sense required for responsibility), there must be an alternate sequence in which there is an action rationalized by his practical reasoning.”Footnote 11 However, in the story of Murder, if Jones had initially refrained from forming an intention to kill Smith and thus had been made to kill him by Black’s intervention, Jones can hardly be said to act as a result of his own character or practical reasoning. In fact, Fischer adds, in such a case it would be hard even to say that Jones is acting at all in the appropriate sense. Thus, claims Fischer, we should not say that Jones is free to do what he does in the appropriate sense, although he is morally responsible for his action (or the particular event of killing Smith). According to him, the fragilist account is flawed because it “confuses the ability deliberately to do otherwise with the [mere] possibility of something different occurring.”Footnote 12
Fischer proffers a new formulation of PAP that better accommodates the appropriate picture of freedom involving alternative possibilities (as he describes it). It goes as follows:
S is responsible for event e only if there [exist] some property F such that F(e) and an alternate sequence open to S in which S brings about ∼F(e´) [(e≠e´)] as a result of an intention to do so (i.e., as a result of an intention to bring about an event with property ∼F).Footnote 13
A couple of clarificatory notes are in order about this formula. First, it concerns one’s responsibility for a particular event resulting from one’s action rather than the action itself. This is because Fischer originally discusses it in his response to a particular version of the fragilist strategy that specifically focuses on events. However, it can easily be revised to cover one’s responsibility for actions as well—e.g., one could refer to a certain action φ-ing and its having or lacking a certain property (F(φ-ing) or ∼F(φ-ing)) instead of an event e, F(e), and ∼F(e). What I will have to say about this formula in my subsequent discussion of Fischer’s argument against the fragilist defense can equally apply to the discussion of the structurally identical principle that concerns our responsibility for actions. For simplicity’s sake, I shall only focus on Fischer’s preceding formula concerning events. Second, in the preceding formula, e´ should not just be regarded as any event distinct from e, though Fischer seems to overlook this point. Rather, e´ is supposed to be some kind of corresponding event of e: it is related in some special way to, though numerically distinct from, e. Roughly speaking, e´ can be characterized as an event that would have occurred “in the place of” e in the relevant alternative scenarios. In other words, e´ occupies the same position in a counterfactual sequence of events that e occupies in the actual sequence: e.g., if e is the immediate result of an agent’s action in the actual world, then e´ will be the immediate result of that agent’s action in a different possible world. To illustrate using the story of Murder, if e is the actual event of Smith’s death caused by Jones’s shooting him via his own will, then e´ will be the counterfactual event of Smith’s death caused by Jones’s shooting him via Black’s intervention.Footnote 14 Bearing these clarifications in mind, let us articulate the aforementioned formula given by Fischer as follows (this articulation being equivalent to his formula):
(PAP*) S is responsible for an event e (where e is the event that occurred in the actual sequence) only if there exists some property F such that
(i) e has F,
(ii) There is an alternative sequence open to S such that e´ is the corresponding event of e in that alternative sequence, and e´ lacks F, and
(iii) In that alternative sequence, S brings it about that e´ lacks F as a result of an intention to do so.Footnote 15
It should be clear that the consequent of PAP* is supposed to represent what Fischer takes to be the appropriate version of the traditional picture of one’s having freedom that involves alternative possibilities. The appropriate version captures the kind of freedom that is robust enough to support the assignment or avoidance of moral responsibility, and since PAP* represents this robust kind of freedom, Fischer argues that incompatibilists who endorse the fragilist account must accept something like PAP*. A problem arises nonetheless for incompatibilists, claims Fischer, because PAP* is undermined by a typical Frankfurt-type case such as Murder. To demonstrate this point, he argues as follows.Footnote 16 Once again, let ‘e’ stand for the particular actual event of killing Smith in the story of Murder, and let ‘e´’ stand for the counterfactual event of killing Smith via the intervention of Black. Also, let ‘F’ stand for the property not being caused by the intervention of Black. It is noted that e has F and e´ lacks F, where e´ is the corresponding event of e occurring in the alternative scenario, but in the alternative sequence where e´ occurs, Jones would not have brought it about that e´ lacks F as a result of his intention to do so. Indeed, a closer look will reveal that in the alternative sequence there is no relevant property that Jones could have brought about as a result of his intention to do so, since in that sequence he would not have been acting through the agency of his own character or practical reasoning―or rather, we should say, he would not have been acting at all in the appropriate sense. For this reason, Fischer concludes that (iii) in the consequent of PAP* is falsified by the story of Murder. As Fischer sees it, this observation gets fragilists into trouble: they are guilty of employing an inappropriate conception of freedom involving alternative possibilities in their typical response to Frankfurt; and yet if they tried to employ a proper notion of freedom involving alternative possibilities, and accepted PAP*, then their account would be vulnerable to Frankfurt-style cases.
3. Defending the Fragilist Account from Fischer’s Critique
Fischer’s new formulation of PAP is illuminating in that it provides a decent starting point for fleshing out the proper notion of freedom involving alternative possibilities associated with moral responsibility. Nevertheless, I believe that there is something wrong in his discussion.
First of all, pace Fischer, the story of Murder as it is told (and other Frankfurt-style cases frequently discussed in the literature on free will) does not undermine PAP*. To demonstrate this point, let us note that Fischer must concede that no property can be found in the story of Murder that satisfies all the three conditions specified in the consequent of PAP*, since he is committed to the claim that PAP* is undermined by the story of Murder. However, when carefully examined, there are such properties. This is because there surely are alternative sequences open to Jones that are consistent with the truth of the consequent of PAP*. To illustrate, consider the following three sequences of events (all of which are consistent with the story of Murder):
(SQ1) Jones decides to kill Smith on his own, and shoots him to death. Black does not intervene.
(SQ2) Jones is inclined to refrain from killing Smith. Black intervenes. By Black’s manipulation of the microchip, Jones shoots him to death.
(SQ3) Jones decides to kill Smith on his own. Black does not intervene. Instead of shooting Smith, Jones slits his throat, causing him to die.
Here, SQ1 is a partial description of what happens in the actual sequence. SQ2 describes what has been referred to as the alternative scenario by Fischer. However, SQ2 need not be regarded as the only alternative, as illustrated by the existence of SQ3, which can also be regarded as a legitimate alternative sequence open to Jones. In fact, there may be many alternative sequences to SQ1—e.g., there may be various ways of killing Smith available to Jones. Now, let ‘F’ represent the property being caused by Jones’s shooting Smith. If we let ‘e’ and ‘e´,’ respectively, stand for the particular event of Smith’s death in SQ1 and the particular event of Smith’s death in SQ3, then the following statements are all true of F:
(i) e has F,
(ii) SQ3 is open to Jones and e´ lacks F, where e´ is the corresponding event of e in SQ3, and thus distinct from e, and
(iii) In SQ3, Jones brings it about that e´ lacks F as a result of his intention to do so.
This observation illustrates that PAP* is not undermined by the story of Murder: Jones is morally responsible for the particular actual event of killing Smith; and yet, Jones could have prevented that particular event from occurring as a result of his own intention in the sense that he could have chosen a different way of killing Smith. This consideration shows that Fischer is mistaken in maintaining that PAP* is vulnerable to the story of Murder. His mistake is to assume that there is only one alternative sequence open to Jones in the story of Murder. As previously illustrated, there are other alternative sequences that contain a property that satisfies the consequent of PAP*, as long as Jones has access to more than one method of killing Smith.
I have shown that Murder is not a counterexample to PAP*, contra Fischer. However, I do not mean to argue that PAP* is immune to every type of Frankfurt-style cases. The type of Frankfurt-style case that does undermine PAP* may be illustrated as follows:
MurderR
Jones killed Smith in the way described in the preceding story of Murder. Black is extremely happy with the way that Jones killed Smith. In fact, if Jones had chosen to kill Smith in a way even slightly different from the way that he in fact killed Smith, Black would have instantly used his power to manipulate the microchip to re-direct Jones’s behaviour.
Unlike the original story of Murder, in this revised story there is only one way of killing Smith available to Jones—namely, by shooting Smith exactly the same way he does in the actual scenario. Each alternative sequence available to Jones is the one where he would have ended up shooting Smith exactly the same way he does in the actual scenario, and yet there he would not have shot Smith as a result of his intention to do so, since he would have been made to do so under the influence of Black. Fischer may point out that, though PAP* is not undermined by cases like Murder, cases like MurderR are clear-cut counterexamples to PAP*.
I am willing to grant that PAP* may be undermined by a limited range of Frankfurt-style cases such as MurderR. However, PAP* can be adequately revised to accommodate cases like MurderR. The key is to revise (iii) in the consequent of PAP*. MurderR falsifies the consequent of PAP* because it falsifies (iii). And it falsifies (iii) because, at the time of killing Smith, there is no alternative sequence where Jones ends up killing Smith differently from the way he actually does without Black’s intervention. Nevertheless, it is still true that Jones could have freely done something, which would have triggered Black’s intervention. Thus, it is tempting to replace (iii) of PAP* with something along this line, and revise PAP* as follows:
(PAP**) S is responsible for an event e (where e is the event that occurred in the actual sequence) only if there exists some property F such that
(i) e has F,
(ii) There is an alternative sequence open to S such that e´ is the corresponding event of e in that alternative sequence, and e´ lacks F, and
(iii´) In that alternative sequence, S does (or is disposed to do) something or refrains (or is disposed to refrain) from doing something as a result of his own character or practical deliberation, which brings it about that e´ lacks F.
Due to the revision in (iii´), PAP** is not vulnerable to MurderR. To demonstrate this point, in the story of MurderR, let ‘e’ and ‘e´,’ respectively, be the particular event of Smith’s death in the actual scenario and the particular event of Smith’s death in the relevant counterfactual scenario where Jones would have been made to kill Smith via Black’s intervention, and let ‘F’ be the property not being caused by Black’s intervention. It is plain that requirements (i) and (ii) are satisfied: e has F while e´ lacks F. Note that, unlike (iii), requirement (iii´) (of PAP**) is satisfied as well: in the story of MurderR, it was up to Jones to cause some event, which would have forced Black to intervene and thereby would have ensured that the corresponding event lacks the property not being caused by Black’s intervention. For instance, Jones, completely out of his own character or practical reasoning, could have aimed, or at least could have been disposed to aim, at Smith’s head rather than aiming at his heart as Jones in fact does in the actual sequence. This piece of action (or this disposition towards acting in a certain way) would have caused Black’s intervention, which makes it the case that the corresponding event would have lacked the aforementioned property. This observation suggests that even the toughest Frankfurt-style cases such as MurderR would not undermine PAP**.
However, PAP** is defective. The problem with PAP** is that the alternative possibilities it allows are not robust enough. To illustrate this point, let us examine more closely how the story of MurderR satisfies (iii´). In the alternative scenario in question, Jones voluntarily could have done something (namely, aiming at Smith’s head), which would have brought it about that e´ (Smith’s death in that alternative scenario) lacks F (not being caused by Black’s intervention). But here Jones’s aiming at Smith’s head causes e´’s lacking F merely as an unintended and unforeseen side-effect. Fischer might claim that the resulting state of affairs in the relevant alternative scenario ought to be rationalizable in terms of the agent’s voluntary intentions in order for that alternative to count as freedom in the robust sense that matters for moral responsibility. Let us call this demand the volitional condition of the robustness requirement.Footnote 17 Furthermore, although Fischer was not explicit in pointing this out, one may argue that in order for an alternative possibility to be robust, the agent should have reasonable grounds to believe that by securing the alternative possibility he would thereby have avoided the responsibility he has for the event he actually brings about.Footnote 18 Let us call this the epistemic condition of the robustness requirement. The challenge then is to formulate a version of PAP that satisfies the two conditions of the robustness requirement, and at the same time, is immune to more resilient Frankfurt-style counterexamples like MurderR. To meet this challenge, I propose the following revision of PAP**:
(PAP*R) S is responsible for an event e (where e is the event that occurred in the actual sequence) only if there exists some property F such that
(i) e has F,
(ii) There is an alternative sequence open to S such that e´ is the corresponding event of e in that alternative sequence, and e´ lacks F,
(iiiR) In that alternative sequence, S brings it about (or is disposed to bring it about) that e´ lacks F as a deliberate result of his intention to bring about e´, and
(iv) There are reasonable grounds for S to believe that S would avoid bringing about e by bringing about e´.
In PAP*R, (iiiR) replaces PAP**’s (iii´) to meet the volitional condition of the robustness requirement. For it states that the agent’s bringing about the resulting state of affairs in the alternative scenario—e´’s lacking F—deliberately stems from his intention to bring about the very event incorporated in that state of affairs (and not as an unintended side-effect). The epistemic condition is met due to (iv): the agent could reasonably believe that by bringing about an event that would preclude the occurrence of the event he actually brings about he would thereby not be responsible for the actual event.
PAP*R is immune to MurderR while satisfying the robustness requirement. To illustrate this point, let us characterize two sequences of events consistent with that story as follows:
(SQ1*) Jones wants to shoot Smith in the heart. He aims at Smith’s heart, and shoots him in the heart. Smith dies. Black does not intervene.
(SQ2*) Jones wants to shoot Smith in the head. He aims at Smith’s head. Black intervenes. Jones shoots Smith in the heart, and Smith dies.
Here, SQ1* is a partial description of what happens in the actual scenario while SQ2* describes a possible course of events where Black intervenes. Now, let ‘e’ stand for the particular event of Smith’s death in SQ1* and ‘e´’ stand for the particular event of Smith’s death in SQ2*. And let ‘F’ stand for the property not following upon Jones’s volition of aiming at Smith’s head. e has F and e´ lacks F—so (i) and (ii) are satisfied. (iiiR) is satisfied as well: in SQ2*, Jones voluntarily intends to bring about e´ (by shooting Smith in the head) as a deliberate result of which he brings it about that e´ lacks F. In other words, the reason why Smith’s death in SQ2* follows upon Jones’s volition of aiming at Smith’s head is because Jones had previously preferred to aim at Smith’s head in the first place—so it satisfies the volitional condition. Also, it is reasonable for Jones to believe that he would not be responsible for Smith’s death in SQ1* if he brings about Smith’s death in SQ2*, for he has a good reason to believe that Smith would be shot in the head in SQ2*. Of course, in both scenarios Smith is (or would be) shot in the heart, so Jones would be mistaken in holding that belief. Nevertheless, it would be still reasonable for him to hold that belief, given that he could not have possibly known about the microchip in his brain. So, the epistemic condition—i.e., (iv)—is satisfied as well.
Fischer argues that the fragilist defense of PAP employs an inadequate notion of freedom involving alternative possibilities; but if fragilists accept PAP*, which he believes incorporates the proper notion of freedom, they would find this principle vulnerable to typical Frankfurt-type cases. In this paper, I have argued that Fischer’s counterargument is unsuccessful. Contrary to Fischer, PAP* is not vulnerable to the most common type of Frankfurt-style counterexamples, such as Murder. Although more resilient counterexamples like MurderR do undermine PAP*, those who wish to defend a plausible version of PAPFootnote 19 can provide a properly revised principle like PAP*R to meet the challenge.
Acknowledgement:
I am grateful to Anthony Brueckner, Sungil Han, John-Michael Kuczynski, Boram Lee, and anonymous referees for Dialogue for their helpful comments and suggestions on earlier drafts of this paper.
