The Frankfurt Counterexample Defended - Nucci

I recently discovered (amazing really, after all this time) the journal Analysis. It is fast becoming one of my favourites. “Why?” you ask. Well, for one simple reason: brevity. Many of the pieces in Analysis are short (under 10 pages), highly relevant and to the point. You have no idea how refreshing this is, particularly for one like me who comes from a legal background where articles can be often over 100 pages (check out US Law Reviews if you don’t believe me).

(Philosophical Trivia: Analysis is the journal in which Edmund Gettier's famous 3-page revolution of 20th century epistemology was published.)

Anyway, I thought I might do a post on the following piece that I recently read in Analysis:

• Nucci, E. D. “Frankfurt Counterexample Defended” (2011) 71 Analysis 102-104.

It covers some material that has already been discussed on this blog.


1. Frankfurt defeats PAP
According to one of the standard positions in the philosophical debate, an agent (X) can only be held responsible for an action (A) if they were able to do something else (~A). In other words, the agent is only responsible for A if they could have done otherwise. This position is captured by something called the principle of alternative possibilities (PAP).

Harry Frankfurt suggested that PAP can be undermined by certain counterexamples, of which the following is one:

Frankfurt Case: Black wants Jones to perform a certain action A. Suppose Black is an amazingly reader of body language cues such that he can tell, in advance, what Jones has decided to do. If Jones decides to perform A, then Black will do nothing; If Jones does not decide to perform A, then Black will intervene and force him to do A. Now imagine that, as it happens, Jones decides to perform A and Black never has to intervene.

Question: Is Jones responsible for A?

Most say “yes”. This seems to create a problem for the defender of PAP because, in the scenario described by Frankfurt, Jones could not have done otherwise. So there appears to be a dilemma: either we accept that Jones is responsible for A and discard PAP, or we retain PAP and discard the belief that Jones is responsible for A.

More formally:

• (1) In the Frankfurt Case, Jones is responsible for A.

• (2) Jones is responsible for A if and only if Jones could have done otherwise (PAP).

• (3) Jones could not have done otherwise in the Frankfurt Case.

• (4) Therefore, either Jones is responsible for A and PAP is false; or PAP is true and Jones is not responsible for A.

2. Larvor Defuses Frankfurt
All is not lost for the defender of PAP. Assuming they do not wish to give up on (1), they can always try to challenge (3). In other words, they can argue that Jones can actually do otherwise in the Frankfurt case because alternative possibilities are available to him.

This is exactly what Larvor argues in a previous article in Analysis. Larvor points out that in the counterfactual scenario, i.e. the scenario in which Black intervenes and forces Jones to perform A, it is not actually the case that Jones performs A. In the counterfactual scenario it is Black who performs A. The fact that he does so through the medium of Jones’s body is incidental.

Larvor then argues if it is not the case that Jones performs A in the counterfactual scenario, then it is the case that Jones can do otherwise in the factual scenario. Why is this? Because Jones actually does face two possibilities in the Frankfurt case: (i) he can perform A of his own volition or (ii) he can get Black to perform A through the medium of his own body.

More formally, we can say that the conjunction of the following two premises defeats (3):

• (5) In the counterfactual scenario, Jones does not perform A, Black does.

• (6) If Jones does not perform A in the counterfactual scenario, then Jones could have done otherwise in the actual scenario.

3. Nucci Defends Frankfurt
We’ve now reached the point in the dialectic at which Nucci’s article actually becomes relevant. Nucci, you see, tries to respond to Larvor’s argument. He does so by drawing a distinction between:

• (a) Not A-ing; and
• (b) Avoiding to A.

He claims that the former does not entail the latter and that this is crucial to the success of Larvor's objection. To put this in slightly less abstract terms: Suppose in the Frankfurt scenario, Black wants Jones to kill a man named Smith. In the actual scenario, Jones decides to kill Smith without any interference from Black. This means, following Larvor, in the counterfactual scenario Jones manages to not kill Smith because Black ends up doing it through the medium of Jones’s body.

Nucci’s point is that not killing Smith is a very different thing from avoiding to kill Smith. Avoiding to kill Smith implies that it is somehow up to Jones whether or not Smith is killed. This is akin to Jones having some kind of power or ability to prevent Smith’s death. Clearly, in the Frankfurt scenario, Jones lacks this ability. After all, Black’s intervention in the counterfactual scenario is not up to Jones; it is something within Black’s control.

How does this save the Frankfurt counterexample from defeat? Well, the idea is that in order for PAP to truly hold, it must be the case that not killing Smith is up to Jones. Since even on Larvor’s interpretation this is not the case, PAP does not hold true in the Frankfurt counterexample. Thus Frankfurt’s original dilemma is preserved.

• (7) To not A is not the same thing as to avoid A-ing.

• (8) To be able to do otherwise is to be able to avoid A-ing.

• (9) In the Frankfurt scenario, Jones cannot avoid A-ing.

So goes it for Nucci’s defence of Frankfurt. I am unsure whether it is successful or not.

Fischer, Frankfurt and Flickers of Freedom (Part 2)

(Part 1)

This post is the second in a brief series looking at John Martin Fischer’s response to the Flicker of Freedom strategy that some have used to undercut the effectiveness of the Frankfurt counterexamples to the principle of alternative possibilities (PAP). Although I assume you have already read part one, some recapitulation of the basic ideas will help us to both frame and better understand Fischer’s response.

The first thing worth recapitulating is PAP itself. As you recall, PAP is intended to describe a necessary condition of moral responsibility. According to PAP, a person cannot be held morally responsible for performing an action unless it was possible for them to do otherwise. Frankfurt-style cases are used as a counterexamples to PAP. These cases describe situations in which agents are unable to do otherwise but still seem, intuitively at any rate, to be morally responsible for what they did. If these cases succeed they are supposed to show that responsibility is legitimate in a deterministic world.

The Flicker of Freedom strategy serves as a counter to the Frankfurt counterexamples. The strategy points out that if you analyse the Frankfurt-style cases closely, you will find that they contain a flicker of freedom: a brief window of opportunity that opens into an alternative possible future before being quickly shut again. Proponents of the strategy argue that the presence of these flickers is what accounts for the intuitive belief that people are morally responsible in these scenarios. Furthermore, this undermines the belief that moral responsibility is legitimate in a deterministic world because if determinism is true then not even flickers of freedom are possible.

We considered four different versions of the strategy last time out. This time we consider Fischer’s actual rebuttal to the strategy.


1. Not Enough Freedom
As noted last time out, Fischer doesn’t think he can offer a decisive rebuttal of the strategy, he does, however think he can offer “a set of considerations” that count against the strategy. Now I know sets can contain only one member, but I find Fischer's comment slightly annoying since, as far as I can make out, only one consideration is offered. But I can forgive that if the consideration is a good one. Is it?

Fischer’s basic contention is the following: the flicker of freedom available in Frankfurt-style cases does not provide an alternative possibility that is sufficiently robust to ground an ascription of responsibility. And since it does not do so, PAP does not explain or justify the intuitive belief that the individuals in these cases are responsible.

The main problem, as Fischer sees it, is that the alternative sequence that flicker-theorists think they have identified would be non-free.

Consider one of our examples: Jones and Black and the decision to vote for the Democrat in the presidential election. In the Frankfurt set-up, Jones votes for the Democrat because that is his wish. However, unbeknownst to him, a nefarious neurosurgeon named Black has implanted a fail-safe device in his brain that will force him to vote Democrat should he show any (neural) indication of voting Republican. In other words, even though Jones is happy to vote Democrat, he could not, as a matter of fact, have done otherwise. The question is whether, in spite of this, he can still be held responsible for what he did. The proponent of Frankfurt-style cases says "yes" and thinks this disproves PAP.

The flicker-strategist also says "yes" but thinks that this does not disprove PAP. The reason being that Jones had the option of showing some indication to vote Republican and could thus bring about his voting for the Democrat via an alternative sequence of events. Fischer’s response is to point out that this alternative sequence would involve a coerced action (a “non-free” action) and to follow-up by pointing out that a coerced action cannot ground responsibility. Thus, he concludes, that the flicker of freedom is not doing the work that the defender of PAP would like it to do.


2. Counterfactual Theories of Knowledge
Fischer improves his response considerably by means of a clever analogy. The analogy is to counterfactual theories of knowledge. Such theories contend that an agent S can only be said to know a proposition P if the agent can “track the truth” of P across multiple possible worlds. In practice, this means that S must be able to distinguish reliably between the possible worlds in which P is true and the possible worlds in which it is false.

Characterised in this manner, counterfactual theories of knowledge have something in common with PAP. After all, PAP maintains that S is responsible for A, if and only if S can bring about worlds in which ~A obtains. Thus, if we can find a problem with counterfactual theories of knowledge that seems similar to Fischer’s objection to the flicker strategy, we might have reason to doubt that strategy.

Fischer thinks such a problem can be found. He asks us to consider an individual, call him Schmidt, who claims to know that there is a barn in front of him. The proponent of the counterfactualist theory will argue that Schmidt can only make this claim if he can distinguish between the actual situation (where there is a barn in front of him) and a range of alternative situations (in which there is no barn in front of him). This is the essence of the counterfactualist theory.

Now suppose for sake of argument that Schmidt cannot distinguish between the actual situation and alternative situations. In other words, suppose that in the alternative situations Schmidt thinks there is a barn in front of him when there is not. This means Schmidt has false beliefs in the alternative situations. Would it not then be odd for a counterfactualist to suggest that it is in virtue of the fact that Schmidt forms false beliefs about the barn in the alternative situations that he has knowledge of the barn in the actual situation? Clearly it would.

But, argues Fischer, such a claim is directly analogous to the claim being made by the proponent of the flicker-strategy in relation to the Frankfurt-style case. They are claiming that it is in virtue of an alternative sequence involving a coerced, non-free action (that they themselves would think is non-responsible) that the individual is responsible for his action in the actual sequence.

This, Fischer suggests, is implausible. And indeed it highlights the true importance of the Frankfurt-style cases. They switch our focus to the actual sequence, and suggest that it is in virtue of the actual sequence, not hypothetical alternative sequences, that agents are responsible. The task of the responsibility-theorist should thus be to identify the features of the actual sequence that make an agent responsible for their actions.

Fischer, Frankfurt and Flickers of Freedom (Part 1)

Frankfurt-style cases are designed to function as counterexamples to the principle of alternative possibilities (PAP). I discussed both of these things in a pair of posts last week. In the comments section to the second of those posts, reader Robert Oerter (of the blog SomewhatAbnormal) brought up an interesting objection to the Frankfurt-style case.

As promised, this series of posts looks at John Martin Fischer’s response to that type of objection.


1. What is the Objection?
First, we must get a general idea of the objection. Consider the following two Frankfurt-style cases:

Case 1: You are taking a driving lesson and you are steering the car around a particularly difficult right hand turn. Unbeknownst to you, your instructor is ready to intervene should you start veering to the left, but is happy to leave you in control if you stick to the right. As it happens, you do stick to the right and your instructor never intervenes.

Case 2: Jones wishes to vote for the Democratic candidate in the forthcoming election. However, unbeknownst to him, Black has implanted in Jones’s brain a specially designed fail-safe device which will force Jones to vote for the Democratic candidate if he shows any sign of voting for the Republican candidate. As it happens, Jones goes ahead and votes for the Democratic candidate.

The second of these is the more traditional version. The idea is that in both cases the agent has morally responsible control over its actions. But in neither case is it true that the agent had the power to do otherwise. This is because of the presence of a counterfactual intervenor (i.e. the fail-safe device or the instructor) which will prevent the agent from doing other than what it did. But if this is true, and if we accept that morally responsible control is present in both instances, then we have effectively denied the relevance of PAP to ascriptions of moral responsibility.

That, at any rate, is how the argument is supposed to run. Robert brought up the following objection to the use of cases 1 and 2 to refute PAP (he was directly commenting on case 1 but his comments can be expanded to cover cases like case 2):

“There seems to me to be a serious problem with this scenario, and with Frankfurt-style cases in general. In the scenario, the instructor will only take control IF YOU ACTUALLY BEGIN STEERING TO THE LEFT. Thus, i[t] is in fact the case that you can stop steering to the right. And it is ONLY in this case that the proposed override becomes operative. 



So, these cases seem to present a contradiction in the way they are treated. Under the given conditions, the intervention only becomes operative if you perform the action ("Stop steering to the right"). Yet the conclusion claims that you CANNOT perform the action.”

Robert’s comment gets at something that John Martin Fischer calls the “Flicker of Freedom”-strategy. This is a strategy that some philosophers have employed in response to Frankfurt-style cases.

The observation that motivates this strategy is the following: in cases like 1 and 2 the agent it not totally compelled to do what they did. They each had the ability to display a flicker of freedom: Jones’ could have displayed some inclination to vote for the Republican and you could have begun veering to the left. And it would only be after those flickers of freedom had been displayed that the counterfactual intervenor would play its part and prevent the agent from doing otherwise.

The indication to vote Republican, which triggers the fail-safe device, is a "flicker of freedom"

Two implications are drawn from this observation:

• (i) The fact that we think the agents in both cases have morally responsible control over their actions is ultimately attributable to the fact that they had the power to display a “flicker of freedom”. This flicker of freedom suggests that alternative possibilities are available to them. And so we need not abandon PAP in thinking they are morally responsible.

• (ii) Frankfurt-style cases have no relevance to the question of whether responsibility is compatible with causal determinism. This is because in a truly deterministic world there would not even be the chance of a flicker of freedom.

Fischer responds to the flicker of freedom-strategy in his essay “Responsibility and Alternative Possibilities” which can be found in his book My Way: Essays on Moral Responsibility. The rest of this series summarises his response.


2. Varieties of the Flicker of Freedom
Fischer begins by describing four different versions of the flicker of freedom strategy. The versions differ in how they identify and conceive of the flicker of freedom that is present in Frankfurt-style cases.


(i) Tracing Back
The first version of the strategy claims that if you trace back the sequence of events in any Frankfurt-style case, you will eventually come across a flicker of freedom: a brief window of opportunity that opens into an alternative possible future. The claim is then that it is impossible to construct an example that lacks at least some form of flicker. And that it is always the presence of this flicker that accounts for our intuitive belief that the agent is responsible in such cases.

For instance, look again at case 1. There is a very obvious flicker of freedom in this case: you can actually start to veer the car to the left, only after you do this will the instructor intervene. In case 2, the flicker is more subtle: the fail-safe device uses some indicator of future decision-making (perhaps a particular pattern of brain activity) before it kicks into action. This indicator is the flicker of freedom. Other Frankfurt-style cases can be constructed in which the flicker is even more subtle, but, so the objection goes, some sort of flicker is always present if you trace back far enough.


(ii) Action-Individuation
The second version of the strategy focuses on how we individuate actions and considers the impact of this process of individuation on our understanding of alternative possibilities. According to a strong version of action-individuation, an action is different if any part of the causal sequence leading up to the performance of the action is different. So, for example, if I pick up my pen with my right hand at the point closest to the tip, I am said to be performing a different action than I would have if I picked it up, with my right hand at the opposite end.

The problem this poses for the Frankfurt-style counterexample to PAP is as follows. Looking at case 2, Jones’s deciding to vote for the Democrat in the actual sequence is a different action from his deciding to vote for the Democrat in the hypothetical sequence involving the fail-safe device. Why? Because in the hypothetical sequence his vote is brought about in a causally distinct manner. It follows that there is the power to realise an alternative possibility in this case. That is to say, although Jones cannot stop himself from voting Democrat, he can bring about his voting for the Democrat in two distinct manners.

Although this analysis applies most obviously when we use a strong version of action individuation, Fischer thinks it could apply equally well for weaker versions because, however you look at it, voting Democrat after some fail-safe device has been triggered does seem to be a different action from voting Democrat without the triggering of that device.


(iii) Agent Causation
The third version of the strategy is quite similar to the second. The difference comes in its appeal to the concept of agent causation. This is a concept that is easy to describe but somewhat difficult to understand. The description is the following: an agent can cause an event in a manner that is distinct from how an event causes another event. All agent causes are free in the sense that they are undetermined by prior events.

This affects our perception of the Frankfurt-style case in the following manner. In the actual sequence, where the fail-safe device does not play a part, Jone’s agent-causes his decision to vote Democrat. In the hypothetical sequence, the decision to vote Democrat is induced by the fail-safe device, not by Jones qua agent. This difference again amounts to a flicker of freedom.


(iv) Doing things on your own
The final version of the flicker-strategy focuses on the type of action the agent is deemed to have performed in the actual sequence vis-a-vis the hypothetical sequence. The argument is that in the actual sequence the agent (e.g. Jones) performs the action “voting for the Democrat on his own”, whereas in the hypothetical sequence he merely “votes for the Democrat”. The difference constitutes an alternative possibility and thus a flicker of freedom.


Fischer says he doesn’t have a knockdown argument against these objections to the use of Frankfurt-style cases. But, then again, knockdown arguments are few and far between in philosophy. Instead, he has a “set of considerations” which lessen the appeal of the flicker-strategy. We’ll look at those in the next post.

Present Luck and Little Agents

(Conceptual Framework on Free Will)

This post is part of my ongoing series on free will and moral responsibility. It goes through a recent (and refreshingly succinct) paper entitled “Lucky agents, big and little: Should size really matter?” by David Blumenfeld. Blumenfeld’s paper challenges Alfred Mele’s alleged solution to something known as the problem of present luck. This is a problem that is said to confront libertarian theories of free will.

The structure of the post is as follows. First, I will introduce and outline the problem of present luck. Second, I will present Mele’s alleged solution to the problem (as it is stated in Blumenfeld’s paper). And third, I will outline Blumenfeld’s critique.

Before I get underway I just want to note that Blumenfeld’s is the only academic paper I have ever read that contains the term “Pussy Pleaser”. Something to think about.


1. The Problem of Present Luck
The claim is as follows: libertarian theories of free will lead to the unpalatable conclusion that whether an individual acts in a particular way (or not) is strictly a matter of luck. This is problematic because responsibility cannot be attached to things that are a matter of luck. So it follows that if libertarianism is true, there are no responsible agents.

We can make this argument more compelling by tracing out the reasoning in more detail. We begin with a definition of the libertarian position on free will. The libertarian position is essentially that the existence of a strong form of free will is necessary for the existence of moral responsibility. This strong form of free will depends on the possibility of basically free actions (BFAs), which are defined in the following manner:

BFA = A is a BFA for S at time t, iff the state of the universe prior to t, coupled with the laws of nature for that universe, are also consistent with S’s not A-ing at t. Or in other words, the prior state of the universe does not necessitate S’s performance of A at t, and is equally compatible with S’s not A-ing at t.

Using this as their foundational requirement, libertarians can adopt one of two positions on the nature of responsibility for individual actions. They can either say that: (i) every responsible action must, in itself, be a BFA or (ii) every responsible action must be capable of being traced back to a BFA.

The second possibility is intended to make allowances for the fact that people could freely decide to, for example, always give money to charity at Christmas at one point in their lives and then bind themselves to that decision at future moments in their lives. So even though their future charitable donations are not basically free, they are traceable back to a prior BFA.

Now we come to the problem of present luck. The problem is arrived at by simply drawing out the implications of the definition of a BFA. I would set it out formally as follows:

• (1) Suppose there is an agent, call him Joe, who performs a BFA (call it “A”) at time t.

• (2) From the definition of a BFA, the prior state of the universe and the laws of nature were consistent with ~A.

• (3) In other words, there was another possible universe, which had the same prior universe-states and the same laws of nature, in which Joe did not perform A at time t.

• (4) Joe’s personality, thoughts, desires, hopes, beliefs, character traits, dispositions etc. are part of the state of the universe prior to t.

• (5) So in the two possible universes under consideration, there is nothing in Joe (i.e. in his personality or mind) that is different.

• (6) This implies that nothing in Joe accounts for the difference between his performing A in one universe and not performing A in the other universe.

• (7) Which is to say: whether A or ~A obtains is strictly a matter of present luck.

In my opinion, this is a rather neat, and highly persuasive piece of argumentation. Indeed, it ranks as, perhaps, the major reason I have for rejecting theories of responsibility that are reliant on libertarian free will.*


2. Mele’s Solution: Little Agents
Blumenfeld’s paper focuses on a proposed solution to the problem of luck that comes from the work of Alfred Mele. Mele is an unusual figure in the debate over free will. He is not committed to either libertarianism or compatibilism and so occasionally offers arguments supporting both sides of the debate. (He is, however, committed to the existence of freedom and responsibility).

In his 2006 book, Free Will and Luck, he takes up the libertarian cause and, after presenting a forceful version of the problem of present luck, he proceeds to offer a frankly bizarre solution to it. The solution is based on the difference between adults ("big agents") and children (“little agents”).

Mele’s basic contention is that libertarians should embrace present luck because its implied by the kind of free will they value. But they can do so unapologetically because responsibility can be built up from actions that are initially due to luck. How so? Well, we can all probably accept that decisions we make early in life can shape the probabilities that attach to our future decisions. That is to say, while our early decisions might be attributable entirely to luck, their consequences might increase the probability with which certain types of decisions are made in the future.

To show how responsibility can get a foothold in this world of lucky decisions, Mele draws our attention to some of the key features of the decisions made by little agents: (i) they are trivial; (ii) the agent has little appreciation for their consequences; (iii) and the agent lacks impulse control. Because of these features, Mele thinks it is okay for the little agent to be held a “bit” responsible for his actions, even though they are attributable to luck. And from this tiny morsel of responsibility for early decisions, can be built a more full-bodied form of responsibility for later decisions that are less attributable to luck (due to the change in probabilities).


3. Blumenfeld’s Riposte
As I said, I find Mele’s argument to be bizarre. I suspect others reading this will have had a similar reaction. Luckily (irony?), Blumenfeld spells out exactly why Mele’s argument is bizarre.

He begins by offering the following formalised version of Mele’s Little Agent Argument (LAA):

• (1) If a little agent makes a BFA and (a) it is relatively trivial; (b) it is much harder for the agent to exercise impulse control over it than it would be for an older child or adult; and (c) the agent is far less able to appreciate its consequences than would be an older child or adult, then the agent has a small degree of moral responsibility for it even though it involves present luck.

• (2) Little agents perform BFAs with the features (a) - (c).

• (3) Hence, little agents can have a small degree of responsibility for their BFAs even though they involve present luck.

Blumenfeld could have added to this the further argumentation to the effect that full responsibility can be arrived at once the small degree of responsibility is compounded into the agent’s maturation process. But since his concern is with this first portion of the argument, I can see why he doesn’t do this.

Anyway, Blumenfeld expresses two main concerns about the LAA. Of lesser significance is his observation that little agents can perform BFAs for which conditions (a) - (c) do not hold and that Mele is silent on the question of their moral responsibility. Of greater significance is Blumenfeld’s observation that there’s no reason to think that features (a) - (c) actually block the problem of present luck.

Blumenfeld makes his point by arguing that if conditions (a) - (c) don’t block the problem of present luck in the adult case, there’s no reason to think they block the problem in the childhood case either.

To see this, go back for a moment to our hypothetical (and adult) agent Joe who had to decide whether to perform A or ~A and, in the end, chose to perform A. As we saw when discussing this example, the libertarian position is that there is nothing in Joe that could account for the difference between his performance of A or ~A.

This implies that no responsibility could be attached to Joe for his performance of A (since responsibility is a feature of the relationship between an agent and his act). This is true for all decisions that are not trivial, whose consequences are known to Joe, and for which Joe does not lack impulse control.

Now ask yourself: if responsibility doesn’t attach to Joe in the situation just described, how could it possibly attach when Joe’s decision has the features (a) - (c)? Note that conditions (b) and (c) are usually factors that reduce or excuse agents from responsibility and so they don’t seem to help make any sort of case for the idea of a small degree of responsibility. Triviality, or condition (a), might be thought to help, but if the decision is still attributable entirely to luck, its hard to see why even this is the case. After all, there’s nothing about Joe that is responsible for even the trivial act.

So, in the end, it seems that premise (1) of the argument is false and with its falsity Mele’s solution to the problem of present luck must be abandoned.


* Proponents of agent causation, like Timothy O’Connor, might have a way of answering the problem of present luck. Agent causation relies on the idea that agents can cause events in a manner that is entirely distinct from ordinary event-event causation. O’Connor’s argument, if I understand him, would be that there is in fact something that accounts for the difference between the two possible universes discussed in the argument outlined above. That something is the “agent”. The plausibility of this argument depends on whether one accepts the conception of an agent (as something distinct from psychological events and processes) employed by agent causationists. I find this conception to be implausible, so I find O’Connor’s response to the problem of luck to be uncompelling. However, I will admit that I have more work to do before I fully understand what the agent causationists are really arguing.

Action, Responsibility and the Ability to do Otherwise (Part 2)

(Part One)

This post is part of a brief series on Justin Capes’s paper “Action, Responsibility and the Ability to do Otherwise”. As we saw last time, Capes’s paper deals with the principle of possible refrainment (PPR) and the principle of alternative possibilities (PAP). Both of these principles feature in the debate over free will and moral responsibility.

According to PAP, in order for a person to be held responsible for something it must be the case that they could have done otherwise; or, what amounts to the same thing, that they could have refrained from doing what they did. According to PPR, in order for a person to perform an action (A), it must be the case that they: (a) have the power to do A; and (b) have the power to refrain from doing A.

PPR is often used to support PAP, sometimes as a premise in an argument for PAP, sometimes as a challenge to the popular Frankfurt counterexamples to PAP. We reviewed both of these uses in part one. In the remainder of this post, we will consider Capes’s objections to PPR.

Obviously, if Capes’s objections are successful, PPR cannot be used to support PAP. This is probably the most significant implication of the following arguments.


1. Action, Control and PPR
Capes begins his critique of PPR by asking why it is that some people think the power to refrain from performing an action is a necessary condition of being able to perform an action. He thinks the attractiveness of PPR comes from the close association between control and action.

The idea is this: according to most accounts, for something to count as one of your actions you must be in control of it. But, so the story goes, you are only in control of something if you can refrain from doing it.

To put this in more formal terms:

• (1) For A to count as one of your actions, you must be in control of A.

• (2) You are only in control of A if you can refrain from A-ing.

• (3) Therefore, A counts as one of your actions iff you can refrain from A-ing.

Of course, (3) is equivalent to PPR. So if you accept the link between control and action, and you accept the definition of control contained in premise (2), you must accept PPR.

But therein lies the rub. As Capes points out, there is no reason to accept the definition of control contained in (2). He uses an example from Fischer and Ravizza to make his point (this is my paraphrase):

Imagine you are taking a driving lesson and you are steering the car around a particularly difficult right hand turn. Unbeknownst to you, your instructor is ready to intervene should you start veering to the left, but is happy to leave you in control if you stick to the right. As it happens, you do stick to the right and your instructor never intervenes.

Now ask yourself the question: are you in control of the car (i.e. are you driving the car)? Fischer, Ravizza and Capes think the answer is clearly “Yes!”. But then think about the nature of that control. In this example, it is not the case that you can stop the car from steering to the right because if you do your instructor will intervene. In other words, you are control despite the fact that you cannot do otherwise. Thus, premise (2) is false.

Fischer and Ravizza explain this by distinguishing between two types of control: (i) guidance control and (ii) regulative control. The latter requires the ability to refrain from doing something; the former does not. Guidance control is a property of the actual sequence of events that leads up to the performance of an action, not a property that extends across multiple possible worlds. As we saw in part one, Fischer thinks it is the actual sequence that matters.


2. Action and OCD
The second of Capes’s objections to PPR at first struck me as being a little strange. It was obviously something that had come up in the literature, but I couldn’t help but think there was an easy solution to it for the proponent of PPR. On reflection, I think it reveals an important point about the distinction between action and responsibility, and the nature of philosophical arguments.

The objection stems from an analysis of the behaviour of individuals with certain compulsive conditions e.g. tourettes or OCD. These conditions appear to prevent these individuals from being able to refrain from performing certain actions. But, so Capes’s objection seems to go, such individuals are still capable of action and thus PPR is false.

In support of this, Capes cites the example of Bob who is a compulsive handwasher. He asks us to look at one handwashing sequence from Bob’s life that occurs at T1. He then seems to argue something like the following (this is my reconstruction of his reasoning):

• (4) Bob’s handwashing at T1 is in accordance with his motivations at T1.

• (5) Controlling one’s behaviour in accordance with one’s occurrent motivations is a defining feature of intentional action.

• (6) So Bob’s handwashing at T1 was an intentional action (from 4 and 5).

• (7) Bob’s OCD causes him to lack the motivation to refrain from washing hands at T1.

• (8) Having the motivation to refrain from doing something is a necessary condition for having the power to refrain from doing something.

• (9) So it is not the case that Bob had the power to refrain from washing his hands at T1 (from 7 and 8).

• (10) Therefore, Bob is capable of intentional action despite the fact that he cannot refrain from what he is doing (from 6 and 9).

What is this argument really saying? Well, first of all, it is suggesting that the attribution of intentional action to Bob seems to be appropriate since his motivations directly caused his behaviour. The hidden assumption (which seems to be borne out in the literature) is that proponents of PPR would be inclined to accept this attribution. But then they must confront the second part of the argument which suggests that it is not the case that Bob could refrain from what he did.

In his article, Capes shows how proponents of PPR try to deny the second part of his argument by constructing convoluted analyses showing how those with conditions such as OCD and tourettes actually do have the power to refrain from doing something despite lacking the motivation to refrain. On the whole, I think Capes’s succeeds in showing that these analyses are implausible.

But this is where my initial confusion arose. Why is it that proponents of PPR must accept the first portion of the argument? Why couldn’t they just use PPR to deny that Bob is capable of action? On reflection, I think there are two things to be said in response.

First, it could be said that this is one of those areas in philosophy where there is a complex interplay of intuitions and principles. The intuitive reaction to cases like Bob, at least according to Capes’s analysis, is that the individuals in question are capable of action. Furthermore, in this instance, those intuitions are thought to trump the principle of possible refrainment.

Second, we can re-emphasise the distinction between action and responsibility. Remember, PPR is supposed to define one of the necessary conditions of action, not one of the necessary conditions of responsibility. So we can say that Bob is capable of action without saying that he is responsible. Indeed, I suspect that it was the conflation of action and responsibility that really led to my initial confusion: I was thinking of how hard determinists might simply say that cases like Bob’s show why no one is responsible for anything. But we're not concerned with that here.


3. Unthinkable Omissions
Capes’s final objection to PPR is, in my opinion, quite interesting. He begins by defining a particular subset of actions known as unthinkable actions:

“Unthinkable Action” = A-ing is unthinkable for S if, at time t, S is unable to intentionally A.

To give this some meaning, imagine you have been taught all your life that torture is always and everywhere a moral wrong. As a result of this teaching, when you are confronted with a situation in which you are ordered to torture another human being you find it impossible to muster up the motivations necessary for carrying out this order. Thus, torture is, for you, an unthinkable act.

(Note: “unthinkable” may be a misnomer since the point is not that you are incapable of thinking about torture, but, rather, that you cannot motivate yourself to perform the act.)

Capes argues that there may be many instances in which actions are unthinkable for agents. Indeed, it may be a distinguishing mark of the morally virtuous agent that he/she in incapable of performing morally impermissible actions.

This is where things get interesting. Capes argues that in addition to the phenomenon of unthinkable acts, there may be the related phenomenon of unthinkable omissions (where an omission is a failure to act). In other words, there may be situations in which an individual is incapable of refraining from acting in particular way.

The example Capes gives concerns a mother who cannot prevent herself from forming the motivations necessary to help her alleviate her own child’s suffering. Such a woman would surely be commended for her efforts in trying to alleviate the child’s suffering.

Note what is going on in this example: the mother is incapable of refraining from performing whatever actionsare necessary for her to help alleviate her child’s suffering. In other words, she violates PPR. But surely we would not deny that such a woman was capable of acting? And if we don’t deny this, we once again cast doubt on PPR.

Action, Responsibility and the Ability to do Otherwise (Part 1)

(Conceptual Framework on Free Will)

This is the first part in a brief series on Justin Capes recent paper “Action, Responsibility and the Ability to do Otherwise”.

Capes paper looks at the relationship between two principles that are popular in the free will debate: (i) the principle of alternative possibilities (PAP) and (ii) the principle of possible refrainment (PPR). PPR is sometimes used to support PAP, but Capes objects to this on the grounds that there is reason to reject PPR.

The ensuing discussion will make little sense if we don’t define our terms, so let’s do that now:

(i) PAP: This principle maintains that a person is morally responsible for an action (call this “A”) iff they could have done other than A; or, in other words, iff they could have refrained from A-ing.

(ii) PPR: This principle maintains that A counts as one of S’s actions iff: (a) S has the power/ability to do A; and (b) S has the power to refrain from A-ing.

Obviously, PAP is about the necessary conditions of moral responsibility while PPR is about the necessary conditions of action. The two are related for the straightforward reason that action is integral to most theories of responsibility.

As Capes notes, PPR can be deployed in two different ways as part of the effort shore-up PAP. First, it can be used as a premise in an argument to support PAP. Second, it can be used as part of an objection to one of the most popular counterexamples to PAP.

We will look at both of these uses in the remainder of this post. Before we do that, however, we should note that both PAP and PPR make reference to the power or ability to “refrain from acting”. If PPR is to be used to support PAP, then this phrase must mean the same thing in both principles. Otherwise, the fallacy of equivocation would be committed.

So what does it mean for an agent to have the power to refrain from acting? There appear to be two aspects to this. First, the agent must possess a general capacity that allows them to refrain from A-ing. And second, that general capacity must be operational at the particular historical moment under consideration.


1. From PPR to PAP
With the preliminaries out of the way, we can now proceed to examine the different ways in which PPR is used to support PAP. Capes begins this process by outlining the following simple argument from PPR to PAP:

• (1) A is something for which S might be responsible if A is among S’s actions.

• (2) A is among S’s actions iff S it is within S’s power to refrain from A-ing (this is PPR).

• (3) Therefore, A is something for which S might be responsible iff S has the power to refrain from A-ing (this is PAP).

Although this is a very straightforward argument, Capes thinks it is flawed for reasons unconnected to PPR. He thinks that premise (1) is obviously false because we are often held responsible for things we omit to do (omissions) or for the downstream consequences of our actions and omissions.

Here’s a simple example of what Capes is talking about. It is well-known that it is a criminal offence to cause death by dangerous driving. But what happens if you are feeling drowsy, fall asleep at the wheel, and your car subsequently injures and kills someone? Surely you can’t be said to be acting at the time at which your car injures and kills the person in question?

There are, in fact, several legal cases dealing with this issue (or something very close to it). In general, you would be found criminally responsible for causing death by dangerous driving. But how can this be if you were not acting at the time the offence was committed? One way of analysing these cases is to argue that your responsibility for causing death is derived from your prior fault of falling asleep at the wheel. Thus, you are responsible not for your actions, but for the downstream consequences of your failure (omission) to stop driving when feeling drowsy.

In light of examples like this, Capes tries to salvage the argument from PPR to PAP by distinguishing between direct and indirect responsibility. Direct responsibility arises when we are held responsible for what we actually did as opposed to what we failed to do or what occurred as a result of what did or failed to do.

This allows Capes to formulate the following argument:

• (1*) S is directly responsible for A, iff A is among S’s actions.

• (2) A is among S’s actions iff it was within S’s power to refrain from A-ing.

• (3) Therefore, S is directly responsible for A iff it was within S’s power to refrain from A-ing.

Having revised the argument by making premise (1*) more plausible, Capes thinks the only question remaining is whether premise (2) is plausible. He thinks not and we will consider his arguments in part two.

Allow me to pause at this stage to offer some commentary of my own. While Capes may think his revised premise (1*) strengthens the argument from PPR to PAP, it is my opinion that it actually substantially weakens the argument. (Capes would probably not be bothered by this since he rejects PAP anyway).

Why so? Well, PAP is generally taken as being a necessary condition for all sorts of responsibility, direct and indirect. Indeed, I can’t see why it would be of much use if it didn’t cover all of these cases. By limiting the scope of the argument to cases of direct responsibility one is tacitly forced to accept that there are other cases in which PAP might fail to be a necessary condition for responsibility. This means that PPR can, at best, provide only partial support for PAP.


2. Frankfurt cases and PPR
The second way in which PPR is used to support PAP is to challenge the conceptual coherence of the most popular objection to PAP, namely, the Frankfurt-style cases.

Frankfurt-style cases are a set of thought experiments, originally conceived of by Harry Frankfurt, that are designed to undermine our intuitive endorsement of PAP. The most famous of these thought experiments involves Jones, a hapless voter in a presidential election, and Black, a nefarious neurosurgeon.

The set-up is as follows: Jones wishes to vote for the Democratic candidate in the forthcoming election. However, unbeknownst to him, Black has implanted in Jones’s brain a specially designed fail-safe device which will force Jones to vote for the Democratic candidate if he shows any sign of voting for the Republican candidate. As it happens, Jones goes ahead and votes for the Democratic candidate.

The question is: given these conditions, is Jones responsible for voting for the Democrat? The suggestion, by Frankfurt and others, is that he is. But if one accepts this suggestion one must reject PAP because the conditions of the thought experiment are such that, due to the presence of the fail-safe device, Jones could not have refrained from voting for the Democrat.

Capes notes that proponents of PPR have challenged the use of Frankfurt-style cases on two grounds.

First, they argue that because the set-up is such that Jones could not have refrained from voting for the Democrat, it follows, given PPR, that Jones did not act when voting. Thus, his casting a vote for the Democrat is not among the class of events for which he can be held directly responsible.

Second, they argue that Frankfurt cases appeal to counterfactual scenarios (i.e. what would have happened if the fail-safe device became operational). In particular, Frankfurt cases rely on the suggestion that Jones would have been forced to perform a particular action in a counterfactual scenario. But, according to proponents of PPR, this is incoherent because it is not clear that something that one is compelled or forced to do can count among one’s actions. And since acting is a necessary precondition of direct responsibility...

John Martin Fischer, the foremost defender of the Frankfurt-style case, has a ready-response to all of this. He argues that the function of the Frankfurt-style case is to show us that responsibility and action are constituted by aspects of the actual sequence of causation leading up to an event, not by the hypothetical counterfactual sequence. In other words, Frankfurt cases are designed switch our focus from what is possible to what is actual.

In relation to action in particular, Fischer asks us to imagine a truncated version of the Frankfurt-style case outlined above in which we are told Jones voted for the Democrat without any mention of fail-safe devices. Fischer argues that we would never deny that Jones acted when voting in this truncated version of the story. But if that’s true, then we have no reason to doubt that he acted in the non-truncated version either. Why? Because in neither version did the fail-safe device play any actual role in what Jones did. In other words, action is a function of what actually happened, not what could have happened.

Capes is basically in agreement with Fischer in this debate. His goal, in the remainder of his article, is to find reasons for doubting PPR. We will look at these next time out.

Who's Still Afraid of Determinism? (Part 3)

(Part One, Part Two)

In the previous entry, we encountered Dennett and Taylor's two main arguments contra the incompatibilists. In this brief final entry, we consider the appendix to D & T's article in which they deal with Peter Van Inwagen's famous Consequence Argument.

The Consequence Argument is supposed to show that agents cannot have meaningful causal powers in a deterministic universe. The argument is of moderate complexity. It is outlined and illustrated in the following diagram (click to "embiggen"):

As you can see, the idea is that because determinism implies that events in the remote past are sufficient for events in the future, no one really has control over those future events.

D & T attack the argument on the grounds that it conflates causal necessity and causal sufficiency. As we saw the last time, D & T argue that counterfactual necessity is the most crucial criterion of causal power. Thus, they think that the term "power to cause" in premise (3) of the Consequence Argument should be thought of in the following terms:

A has the power to cause a iff for some sentence ψ describing an action of A and a world f close to actuality, ψ ∧ a holds in f and a → ψ in every world similar to f.

What this means is that within a cluster of possible worlds close to our own, there is an action ψ that is a necessary condition for a to occur. If this sounds confusing or imprecise, I suggest going back to part one and looking at the definitions of possibility and causation that were outlined there.

The problem for Van Inwagen is that once this definition of causal power is employed, premise (3) as a whole becomes unwarranted. Why so? Well, premise (3) in its original form is claiming that a → ψ applies in a cluster of nearby worlds, and that a → b applies in all possible worlds. If we could deduce from these two claims that b → ψ then premise (3) would be justified. But we can't do this because elementary logic tells us that a → ψ and a → b do not entail that b → ψ.

And so, with this simple combination of definitional judo and elementary logic, the Consequence Argument fails. I should say, before leaving D & T's article behind, that many doubt that Van Inwagen's argument can be defeated in such a simple manner (D & T acknolwedge this fact in a footnote). I personally prefer Gary Drescher's suggestion that a causal link between practical reason is not always required for meaningful action. Drescher's argument is presented in his book Good and Real and in effect defangs the conclusion of the Consequence Argument.

Who's Still Afraid of Determinism? (Part 2)

(Part One)

In the previous entry we took a look at the motivations behind D & T’s article and outlined some of the formal concepts they use in making their arguments. In this entry, we turn to the arguments themselves.

These arguments respond directly to the alleged problems with determinism. As you recall, these were (a) that determinism rules out any meaningful sense of alternative possible futures; and (b) that determinism rules out any meaningful sense of originative causal powers.

Let’s see why D & T think those concerns are misplaced.


1. Determinism and Alternative Possible Futures
Dennett thinks that the following famous footnote in one of J.L. Austin’s essays sums up the incompatibilist position on alternative possible futures:

"Consider the case where I miss a very short putt and kick myself because I could have holed it. It is not that I should have holed if I had tried: I did try, and missed. It is not that I should have holed it if conditions had been different: that might of course be so, but I am talking about conditions as they precisely were, and asserting that I could have holed it."

In this short paragraph, Austin is making a claim about what it is possible for him to do. In the previous entry, we saw that such statements are characterised by their appeal to possible worlds. So, roughly, what Austin is saying here is that the sentence “Austin holes his putt” is true in at least one possible world within a specified subset (X) of possible worlds.

The big problem for Austin is how large the specified subset X needs to be in order for him to meaningfully claim that he “could have holed the putt”. In the portion quoted above, Austin seems to include just one world in the subset X: the actual world in which we live (“conditions as they precisely were”). D & T call this the “narrow” approach and think it is typical of the incompatibilist position.

D & T’s contention is that the narrow approach is flawed: to make meaningful claims about what is possible, we have to have a broader interpretation of the subset X. Indeed, Austin himself seems to be aware of this because after the section just quoted he continues in the following manner:

“Further experiments may confirm my belief that I could have done it that time, although I did not.”

If Austin sticks to the narrow approach, “further experiments” are meaningless since they occur in a different set of conditions. Only if he takes the broader approach would such a claim make sense.*

D & T reinforce this observation with a lengthy thought-experiment about two chess-playing computers. Anyone who has spent time reading Dennett will know that, along with Conway’s Game of Life, this is one of his favourite “intuition pumps”. Apparently, there is no concept in the philosophy of mind, biology and responsibility that fails to be illuminated by it.

I won’t repeat the lengthy discussion of the chess-playing computer programmes here. Suffice to say, the important point about such programmes is that they are “miracles of determinism”: they play chess games in accordance with predetermined algorithms. Nevertheless, they can they can play different sequences of games if they make use of pseudo-random number-generators. Each one of those games is a possible world (in a subset X), and each one reveals something about the competencies and capabilities of the computer programme. Or, to put it another way, the games reveal something about what it is possible for the programmes to do. This is a meaningful sense of possibility, and it is based upon a broad construal of the set X.

D & T argue that these observations make discussions of possibility perfectly compatible with a deterministic worldview. The ball is thus sent back to the incompatibilist’s court: they must show why indeterminism is necessary.


2. Making a Difference in a Deterministic World
The second worry that incompatibilists have about determinism relates to causal powers. To recap, the concern is that determinism leads inexorably to the view that human beings are mere receptacles through which non-human causal forces exert their powers. D & T obviously think this is wrong and they do so on the basis of a simple argument:

• (1) Determinism is a thesis about causal sufficiency, not a thesis about causal necessity.

• (2) For meaningful, morally relevant causal powers, we care about causal necessity, not sufficiency.

• (3) So, the truth of determinism does not remove or eliminate the kinds of morally relevant causality that we care about.

Premise (1) could be taken as a stipulative definition of what determinism is, however, that may be a little unfair since it also tracks well what I have read elsewhere. D & T say the definition amounts to something like the following:

If σ1 is a (mind-bogglingly complex) sentence that specifies the state of the universe at t1, and σ2 is another sentence that specifies the state of the universe at t2, then determinism dictates that σ1 is sufficient for σ2 in all physically possible worlds.

What does this definition actually mean when making a causal judgment with practical implications?

D & T look to the example of JFK’s assassination by Lee Harvey Oswald. According to determinism, the condition of the universe one instant after the big bang (call this “σ”) sufficed to produce the death of JFK in November 1963 (call this “ψ”). But, argue D & T, this tells us nothing about what actually caused Kennedy’s death, or who bears responsibility for his assassination.

This is because although σ is sufficient for JFK’s assassination, it is not necessary. JFK could have been assassinated even if σ were other than what it was. The necessary preconditions of the assassination included things like “Lee Harvey Oswald was born” and “Oswald pulled the trigger of his gun” and so on, not the complete description of the microphysical properties of the universe one instant after the big bang.

This example allows us to see why the incompatibilist fear of determinism is misplaced: when we are interested in explaining why something happened, or in figuring out who bears responsibility for it, we are primarily interested in what is causally necessary, not what is causally sufficient. This is what is asserted in premise 2. The conclusion follows as a matter of course.

This brings us to the end of D & T’s main article. They have now, to their minds, shown why there is nothing to fear from determinism and why incompatibilists bear the burden of proving otherwise. They have an appendix in which they deal with Peter Van Inwagen’s famous Consequence Argument. I’ll look at that in the next entry.


* I sometimes like to employ a explicability argument in response to those who make use of the narrow approach. As follows: Suppose one does accept the narrow approach, and suppose the one world within the set X does give rise to alternative possible futures. Ask yourself: what is making the difference in realising these possible futures? By stipulation, it can’t be anything in your personality or psychology because if they were different, conditions would not be “precisely as they were”. So what could it be? The answer: nothing that is relevant to an ascription of responsibility. Is this a good argument? I think so, but I'd be interested in hearing counterarguments.

Who's Still Afraid of Determinism? (Part 1)

Daniel Dennett

(Conceptual framework for understanding free will)

“Who’s still afraid of determinism?” is an article by Dan Dennett and Christopher Taylor (hereafter “D & T”) on the topic of free will and determinism. As the adverbial use of the word “still” in the title suggests, it is an updated version of an earlier article. In the next two blog posts I want to summarise this paper.

Having read both versions, I can tell you now that the difference between them is slight. The major change is that a section has been added discussing Judea Pearl’s theory of causation. Ironically, I’m going to exclude this from my summary. I do so for two reasons: (i) the value of the discussion is proportional to one’s understanding of Pearl's rather complex theory of causation, which I have no intention of summarising and (ii) the authors argue that Pearl’s theory is not relevant to their overall conclusions so, at least in terms of understanding those conclusions, we do not need to discuss it here.

Anyway, my consideration of D & T’s article will be broken into three parts. The first part goes through the motivations behind the article, as well as the various concepts that D & T employ in their arguments. The second part examines the two main arguments (or theses) put forward by the authors. The third part looks at Van Inwagen's famous Consequence Argument, as well as D & T's responses.


1. The Motivations Behind Incompatibilism
The target of D & T’s argumentative ire is incompatibilism. As noted on a previous occasion, incompatibilism is the claim that either (a) free will is incompatible with causal determinism or (b) causal determinism is incompatible with moral responsibility. D & T’s primary concern is with the moral and existential implications of determinism. It would be a mistake, I think, to interpret them as making claims about the ontological or metaphysical implications of determinism.

As D & T point out, incompatibilism is attractive for two main reasons.

First, most people think that in order for alternative possible futures to be available to them when they act, causal determinism must be false. In other words, they think that determinism implies that they “could not have done otherwise”. This is because, in the words of Van Inwagen, determinism is the thesis that there is only one possible future.

Second, most people think that causal determinism implies that human beings are the empty receptacles of external causal forces. This thereby robs them of any ability to be the true (ultimate) originators of their actions.

D & T’s goal in their article is to show how acceptance of causal determinism neither rules out a meaningful sense of “could have done otherwise”, nor a meaningful sense of causal origination. If they succeed, the burden of proof then shifts to the proponents of incompatibilism. They must explain why their position still remains attractive.

To make their case, D & T begin by setting out some formal concepts.


2. Possible Worlds
The first thing D & T try to do is to get some purchase on the modal concepts of possibility and necessity. Philosophers usually do this by appealing to the idea of different possible worlds. D & T are no different, but they focus on possible worlds that are consistent with a modern scientific view of reality. They do not focus on all logically or metaphysically possible worlds.

Their account builds on a Democritean model of the universe. A Democritean universe is specified by a function ( f ) that assigns a value (0 or 1) to a four-dimensional unit of spacetime (x, y, z, t). If a value of 1 is assigned to (x, y, z) at time t1, then that unit of spacetime is occupied; if a value of 0 is applied to that unit, then it is not occupied.

This brings us to a definition:

• Def. “Possible world” = Any function of the form f : (x, y, z, t) → {0,1}

There are two important sets of such possible worlds:

• Ω is the set of all possible worlds.

• Φ is the set of physically possible worlds in which no physical law is broken (i.e. the deterministic worlds).

D & T acknowledge that this Democritean vision is a long way from the universe as described by modern physics, but they think it will suffice as we try to refine our concepts of necessity and possibility.

Within a possible world there will be entities. These are simply connected hypersolids occupying units of spacetime that have coherent and stable clusters of properties. A system of informal predicates such as “is human” and “is alive” will be used to describe such entities.

Some of the predicates ascribe necessary or contingent properties to entities. For example, a sentence of the form “Necessarily, Socrates is mortal” ascribes a necessary property to the entity we call “Socrates”. What does such a sentence really mean? According to D & T, it means roughly the following:

• In every world f, the sentence “∀x (x is Socrates → x is mortal)” obtains.

The crucial question for discussions of necessity is: how large should the set f be allowed to range? Should it cover Ω, or Φ, or some more restricted set X? D & T think this is a difficult question to answer but they propose that the following notation be used to indicate whenever something is necessary:

• □x ψ

This translates as “the sentence ψ obtains in all worlds in the set X” and would be equivalent to “Necessarily ψ.”

Having developed this understanding of the concept of necessity, the concept of possibility is much easier to define. Take the sentence “Possibly, Socrates has red hair”. This sentence can be rendered formally:

• There exists (within the set X) a possible world f in which the sentence “∃x (x is Socrates ∧ x has red hair) obtains.

More generally, we can employ the following notation to indicate that something is possible:

• ◊ x ψ

This translates as “the sentence ψ obtains in at least one world within the set X” or "Possibly ψ".

3. Counterfactuals
Counterfactuals are propositions that straddle possible worlds. As noted in another post on this blog, they are often thought to be central to accounts of causation. The following statement is a classic instance of a counterfactual:

• “If you had tripped Arthur, he would have fallen”

This means, roughly, that in every possible world within the set X, whenever the antecedent (Arthur’s being tripped) holds, so does the consequent (Arthur’s falling). D & T employ the following notation for counterfactuals:

• □ x ψ → φ

In this case ψ means “Arthur was tripped” and φ means “Arthur fell”. Two other bits of notation need to added at this stage:

• "ψ□→ φ"  and "~φ□→ ~ψ"

The first of these translates as “If ψ, then φ”. The second translates as “If not-φ, then not-ψ”.

Again, crucial to understanding the scope of counterfactual statements like this is the question: how large should the set X be construed? The answer, according to D & T is that X ought to include:

• (i) worlds in which ψ holds, not-ψ holds, φ holds and not-φ holds.

• (ii) worlds that are otherwise very similar to the actual world (i.e. the one in which we live).

Obviously (ii) is a bit of a fudge but D & T argue that a certain vagueness is inherent in even the most rigorous discussions of counterfactuals (they reference their later discussion of Pearl’s theory in this regard).


4. Causation
We come at last to the nature of causation. D & T do not think that a “true” account of causation is possible, so they set their sights a bit lower. They try to develop some conceptual tools for refining our thinking about causation in the real world. We will describe these tools by referring to the following example:

• “Betty’s tripping Arthur (call this “ψ”) caused Arthur to fall (call this “φ”).”

Our goal now is to describe the concepts that allow us to make causal claims like this.

• (a) Causal Necessity: ψ is a necessary causal condition for φ if, in every possible world within the set X, whenever ψ does not happen, φ does not happen either. As it happens, ψ is not a necessary condition for φ: Betty’s tripping Arthur is not necessary for his falling, he could fall for other reasons.

• (b) Causal Sufficiency: ψ is a sufficient condition for φ if, in every possible world within the set X, whenever ψ happens, φ happens. In our example, ψ probably is a sufficient condition for φ: Betty’s tripping Arthur probably does make him fall (but perhaps there are ways in which he can avoid this outcome even after he is tripped).

• (c) Independence: in making a causal claim, we expect the two sentences ψ and φ to be logically independent. In other words, we expect there to be possible worlds, however remote they might be from our own, in which ψ is true but φ is not.

• (d) Temporal Priority: in order for one event to cause another, it is usually the case that the cause precedes the effect.

• (e) Miscellaneous other concepts: there are other concepts to which we appeal when making causal claims. For example, most of the time the cause is active while the object on which it exerts its causal powers is passive. Also, physical contact is often thought to be important in discussions of causation.

After describing these conceptual tools D & T proceed to consider three classic problem cases. I won’t go through these here but I will summarise the main points, which are:

• (i) When trying to identify “the” cause of an event, sometimes we focus on the necessary conditions, sometimes on the sufficient conditions, and sometimes on other conditions such as temporal priority.

• (ii) Sometimes there may be no singular “cause” of an event.

Okay that’s enough for now. There's quite a lot to take in here. In the next post we’ll see how these concepts are used in the service of D & T’s arguments. You might like to keep this post open in a separate tab when you read the next part.

Free Will: A Conceptual Framework

Free will, if it exists, is a property of agency. It is something that agents, in virtue of their constitution, can exhibit that non-agents cannot. Furthermore, free will may be the most morally, spiritually, and existentially important property of human agency.

It has occurred to me that I might like to look at some recent papers on the topic of free will on this blog. Those papers tend to assume that the reader is familiar with the ins and outs of the contemporary debate on this issue. I don’t like to make those kinds of assumptions, partly because you never know who might be reading a blog, and partly because reacquainting oneself with the basics of an issue is always worthwhile.

This post offers a conceptual framework for analysing the contemporary debate on free will. The framework comes in three sections. The first section examines the nature of free will as a property of agency; the second section considers the intellectual significance of the debate; and the third section outlines some of the positions one can take up in this debate.


1. The Nature of Free Will
No one would deny that the term “free will” is ambiguous. A lot of conceptual baggage has been attached to those two simple words over the years. This is one reason why the debate over free will (even in the philosophical literature) can be so frustrating: different authors apply different meanings to the term and often end up talking past on another.

In an effort to cut through some of that confusion, I like to appeal to a model of free will that I first came across in Henrik Walter’s book The Neurophilosophy of Free Will (he took it from somewhere else). Walter’s contention is that when we talk about the property of free will, we are talking about a decision-making capacity with three components:

• (i) Alternativism: this is the capacity to (meaningfully) choose between different possible futures. In other words, if X must choose whether to eat an apple or an orange, and if X chooses the orange, it must still be possible for X to choose the apple.

• (ii) Intelligibility: this is the capacity to act from intelligible reasons. In other words, X does not simply choose among possibilities at random, X chooses in accordance with reasons, intentions, desires and beliefs.

• (iii) Origination: this is the capacity to be the originator of actions. In other words, X is not simply a passive receptacle through which external causal forces exert themselves but is, in some sense, the active originator of causal forces.

There are two main advantages to thinking about free will in this way. First, by focusing on three elements, this model helps to avoid the pitfalls associated with thinking about only one of the elements. For example, most discussions of free will are preoccupied with the concept of alternativism. But a popular objection to this preoccupation is that an agent with alternativism and nothing else might amount to little more than a random choice-generator. This would not be the kind of morally salient choice with which we are concerned. The extra ingredients of intelligibility and originations are needed for that.

Second, this model is flexible enough to encompass the diversity of positions that exist on the nature of free will. The flexibility stems from the fact that each of the three components can be subjected to strong, moderate or weak interpretations.

For example, a strong version of alternativism might contend that the agent must have been able to realise different possible futures in the exact same circumstances as obtained at the moment of their original decision. A weaker version might argue that sensitivity to changes in circumstances is all that is required. In future entries we will consider the respective merits of such interpretations.

Because one can have different interpretations of the three components, one can think of this model as describing three dimensions along which different theories of free will can vary. It might be the case that weak interpretations do not deserve the label “free will”, but this is something that can be worked out after the different positions have been described.


2. Intellectual Significance
Why do people bother writing and debating the concept of free will? What’s at stake in this debate? I suggest that there are three separate issues to worry about (I think I’m taking this from something Patricia Churchland said, but I can’t be too sure):

• (i) The Metaphysical/Ontological Issue: this is concerned with the reality or non-reality of the different conceptions of free will. The most widely debated ontological issue is the impact of causal determinism on the possibility of free will.

• (ii) The Moral Issue: this is concerned with the type of free will that is necessary for moral responsibility.

• (iii) The Existential Issue: this is concerned with the existential impact of the different metaphysical conceptions of free will. For example, one might ask: if all our choices are causally determined, is practical reason somehow futile or meaningless?

Discussions of free will tend to blend these issues in different ways. This is understandable since how you resolve one of them will affect how you resolve the others. Nonetheless, it is worth keeping them distinct at the outset.


3. Different Positions on Free Will
After over two thousand years of sustained philosophical debate, one can imagine that numerous stances and positions have been identified on both the nature of free will and the moral and existential issues associated with it. It would be difficult to do justice to all of these positions, but thankfully most of the conversation tends to gravitate towards the following:

• Libertarianism: this is the view that strong interpretations of all three components of the will are needed in order for there to be a meaningful sense of free will.

• Incompatibilism: this is the view (usually associated with libertarianism) that either (a) free will is incompatible with causal determinism or (b) determinism is incompatible with moral responsibility.

• Hard incompatibilism/determinism: this is similar to the above. The addition of the adjective “hard” connotes a positive endorsement of the fact that the world is causally determined and that moral responsibility is impossible. Regular incompatibilists tend not to have that positive endorsement.*

• Compatibilism: this is the view that either (a) free will is compatible with determinism or (b) moral responsibility is compatible with determinism. In order to avoid confusion, John Martin Fischer suggests that the latter position be termed semi-compatibilism.

• Agent Causation: this view is a little harder to characterise. Put most simply, agent causationists are primarily concerned with the origination component of free will. They argue that agents cause their actions (or will their actions) in a manner that is distinct from ordinary event-event causation. In other words, they argue that agents are exempt from ordinary causal processes.

So there you have it, a conceptual framework for discussing free will. I will refer back to this post in future entries on this issue.


* "Positive" is meant here in the sense of "believes it to be true" and not "believes it to be a good or desirable thing".