In these last few posts I have been defending my argument that, even if time travel is possible, a time traveler would not be able to commit “auto-infanticide”. In my last post I warned that confusing counterfactual with indicative conditionals can muddle our thinking about time travel. In this post I offer, as a case in point, Ted Sider’s criticisms[i] of Paul Horwich[ii] and me.
Sider says that I argue "in effect" in the following way: If time travel were possible, then "counterfactuals of co-incidence" like the following would be true: "If many, many time travelers went back in time intending to kill their earlier selves, equipped with deadly weapons, hardened hearts, and excellent information about their targets, there would be a long string of co-incidences: slips on banana peels, sudden attacks of remorse, mistaken identities, and so on." A time traveler can kill her earlier self only if these "counterfactuals of co-incidence" are not true. Therefore, no time traveler can kill her earlier self.
But my argument mentions neither actual nor counterfactual co-incidences, nor does it say anything about the failure of "many, many time travelers" to murder their younger selves. Why would Sider think that this is "in effect" my argument?
The answer lies in a thought experiment due to Paul Horwich. Sider thinks that the thought experiment is the source of intuitions about the impossibility of time travel, and Sider thinks that anyone who thinks that time travelers cannot kill babies is also committed, whether or not they realize it, to an argument against the possibility of time travel.
Horwich's thought experiment goes something like this:[iii]
Suppose that time travel has been invented and that a confused consortium of philosophers establishes the Institute for Changing the Past and embarks on a grand-scale experiment for the purpose of proving that the past can be changed (and in significant, noticeable ways). Because babies are easy to kill, and because the effects of killing are permanent (that is, the laws of physics don't permit resurrection from the dead), the Institute decides to focus its efforts on babies who are known to have survived infancy.[iv] Some of the babies they target are babies who grew up to be famous or infamous historical figures -- Napoleon, Stalin, Hitler, Picasso, JFK. Others are ordinary folk, known only to their friends and family. And still others are the baby selves of the time travelers sent out to kill them. The mission of each time traveler is to kill the particular baby who is her assigned victim, thereby providing experimental proof for the claim that the past has been changed, and, therefore, can be changed.
We know the outcome of the experiment in advance; the attempts will all fail. The reason is not mysterious. They will fail because they did fail. And so we can say, of each time traveler, and of each of her assigned targets: if this time traveler tries to kill this baby, she will fail. This is an indicative conditional, and we evaluate it the way we evaluate the conditional about Oswald and Kennedy. It doesn't matter how many time travelers there are, or how many assigned targets. The indicative conditional comes out true, every time, and for the same reasons. And since we also know that every attempt, by every person whatsoever, to kill each of these babies failed, we also know that this indicative generalization is true: For every time traveler, and every baby on the target list, if the time traveler tried (tries) to kill the baby, the time traveler failed (will fail).
More generally, we know that the experiment of the Institute for Changing the Past is based on a misunderstanding about the nature of time travel. Since there is only one actual past, it will occur in exactly the way that it did occur. The targeted babies all lived to be adults, so all attempts, whether by time travelers or anyone else, to kill them will fail. But it doesn't follow that the babies were somehow protected or that the time travelers were somehow disabled; it doesn't follow that the time travelers could not have killed them. It is true, of many pairs of babies and time travelers, that this time traveler could have killed that baby; it just so happened that he didn't. It's not true that if he had tried again, he would have failed again. The failure to kill was not unavoidable; it was just bad luck.
For the reasons I have already given, it is different for those time traveler-baby pairs that consist of a person and her younger self. In those cases, the failure to kill was no accident. In those cases, the counterfactual, as well as the indicative, is true. If the time traveler had tried (again), he would have failed (again). And because this is true for every time traveler-baby pair where the two individuals consist of a time traveler and her younger self, the following counterfactual generalization is also true: For every time traveler, and every baby who is that time traveler's younger self, if the time traveler had tried to kill the baby, the time traveler would have failed.[v]
Horwich doesn't use his thought experiment to argue that time travel is impossible or that time travelers cannot kill the babies who are their assigned targets. However, he does think that it provides the makings for an argument for the conclusion that time travel to the recent past will not occur. His argument is based on the claim that if a large number of time travelers try to change the past, there will occur a pattern of events that is a certain kind of co-incidence and on the claim that we have good empirical evidence that this kind of co-incidence never happens.
Of course, we all know that co-incidences sometimes happen. A coin might come up heads 100 times in a row, and still be a fair coin. Someone might repeatedly fail to walk on a particular tile in the middle of her kitchen floor even though nothing prevents her from doing so. But ordinarily, over the long run, the law of averages kicks in, and a fair coin comes up heads 50% of the time. And, ordinarily, if someone walks about long enough in her kitchen she will, if she is a normal person in a normal kitchen, eventually step on the title that she missed, for no particular reason, the first one hundred times. But in the time travel case, things appear to be different. The Institute for Changing the Past has a large budget and a great deal of patience; they repeat the experiment hundreds, thousands, tens of thousands of times. And every single time something happens to prevent the bullet from killing the baby. I have explained that it is no mystery that the Institute's experiment is a failure. But many people, including Horwich, remain bothered by the fact that such a large number of able-bodied time travelers repeatedly fail to kill unprotected babies. Why do the time travelers keep failing?
Of course, each individual failure has a perfectly good causal explanation. There is no a priori reason to suppose that the explanations are anything out of the ordinary, so we may suppose, as defenders of time travel typically do suppose, that each time traveler's attempt is thwarted by some mundane event: the baby moved, the time traveler slipped on a banana peel, the bullet jammed, a bird flew in the path of the bullet. And so on. But why does a thwarting event always happen? Why are attempts by time travelers to kill babies always correlated with thwarting events? Here the answer seems to be: no reason, it was just an accident, just a co-incidence. But, says Horwich, there are no long-run accidents or cosmic co-incidences and this is due to a contingent but deep fact about the initial conditions of the universe.[vi] Therefore, he concludes, we have good empirical reason to believe that time travel to the recent past will never happen, even though it is logically and perhaps also physically possible.
Sider criticizes Horwich's argument, but his main interest is in using Horwich's thought experiment for the purpose of constructing and criticizing several bad arguments for the claim that time travelers are unfree or that time travel is impossible.[vii] The arguments that Sider criticizes are, indeed, bad arguments. But they are very different from my argument. They are also different from Horwich's argument. What's going on?
Horwich formulates his argument in terms of indicative conditionals. That’s all he needs, since his aim is to argue that we have good empirical reason to believe that time travel to the recent past (while logically and perhaps physically possible) will not happen. "Counterfactuals of co-incidence" play no role in Horwich's argument.
Sider's three bad arguments, by contrast, are all formulated in terms of counterfactuals. The key premise, of each argument, is that time travel entails "counterfactuals of co-incidence":
"If many many time travelers went back in time, intending to kill their earlier selves, equipped with deadly weapons, hardened hearts and excellent information about their targets, there would be a long string of co-incidences: slips on banana peels, sudden attacks of remorse, mistaken identities, and so on."
(In this passage, Sider restricts his attention to the case of time travelers trying to kill their own younger selves, but recall that Sider thinks that auto-infanticide is just a special case of the more general problem of the time traveler's ability to do other things she didn't actually do.)
In the bad argument that Sider attributes to me, there is an additional premise that says that a time traveler is able to kill her earlier self (Baby Hitler, etc.) only if these "counterfactuals of co-incidence" are not true.
First, insofar as Sider uses these "counterfactuals of co-incidence" as a premise in an argument for the conclusion that time travelers are unable to kill their baby selves (Baby Hitler, the baby next door, etc.), Sider's counterfactuals are not the relevant ones. Consider someone in a maximum security prison, and ask whether that person can kill a baby living on the other side of town. We don't answer this question by considering what would have happened if the person had escaped from prison and traveled across town, equipped with deadly weapons, etc., intending to kill that baby! We answer by asking what would have happened if he had tried to get across town to kill that baby. And the answer is 'no', because if he had tried, he would have failed at the first step; the prison walls would have stopped him.
Second, and more important. Why does Sider think it legitimate to replace Horwich's "indicatives of co-incidence" with "counterfactuals of co-incidence"? We are not ordinarily entitled to infer, from the truth of an indicative, to the truth of the corresponding counterfactual, and Sider gives us no reason to think we are entitled to infer from 'if many many time traveler go back in time..., there will be a long string of co-incidences' to 'if many many time travelers went back in time...there would be a long string of co-incidences." On the contrary, we have reason to suppose that "counterfactuals of co-incidence" are false, at least for the general case of time travelers trying to change the past (eg. by killing babies before the day of their death). For if it really is just a co-incidence that all actual attempts by time travelers to kill their targeted babies have been thwarted by banana peels, etc., we have no grounds for supposing that counterfactual attempts would also fail due to thwarts. Only lawlike generalizations sustain counterfactuals; accidental generalizations do not.
This discussion of Horwich's thought experiment and argument, and of Sider's (unconscious?) permutation of Horwich's argument highlights the importance of distinguishing between indicative and counterfactual conditionals, as well as the importance of distinguishing between claims about ways in which a time travel world would be strange and different from the way we believe our world to be (which is what Horwich and I are concerned with) and arguments for the impossibility of time travel.
[ii] Horwich, "On Some Alleged Paradoxes of Time Travel", Journal of Philosophy 72 (1975): 432-444.
[iii] I am adapting Horwich's thought experiment somewhat for my own purposes, but I don't think either Horwich or Sider would object. Horwich wouldn't object because he describes his experiment in terms of "bilking attempts" and he defines bilking attempts as attempts to bring about some past event that did not occur, such as killing one's infant self or doing something one remembers was not done. (p.120). Sider would not object because he views attempts to kill one's infant self as merely an "especially vivid example" of attempts to "do something that did not in fact occur". (p.1)
[iv] So far as changing the past is concerned, it doesn't matter whether death is permanent or not. Even if the world is such that people are constantly rising up from the dead, the past would be changed if it were true that a person is both killed on a particular day and is not killed on that day. I make this point to draw attention to the fact that while we usually think of killing as permanent, this is due to a contingent feature of the world, and not true by definition. (Contrast Sider, p. 1.)
[v] The points made in the last three paragraph are mine, but I do not think that Horwich would disagree with what I say in the first two of these paragraphs (at least).
[vi] Horwich, Asymmetries in Time. http://www.amazon.com/Asymmetries-Time-Problems-Philosophy-Bradford/dp/0262580888/ref=sr_1_8?s=books&ie=UTF8&qid=1318042966&sr=1-8
[vii] The first argument runs thus: time travel entails 'counterfactuals of co-incidence'; these counterfactuals are never true; so time travel is impossible. The second is the argument that he attributes to me: time travel entails "counterfactuals of co-incidence"; these counterfactuals entail the inability of time travelers to kill their younger selves (or Baby Hitler, or any other baby before the date of that baby's death). The third is another argument for the conclusion that time travelers are unable to kill their younger selves: time travel entails either "counterfactuals of coincidence" or "strange shackles"; counterfactuals of co-incidence are never true; therefore time travel entails 'strange shackles'; therefore time travelers are unable to kill their baby selves (or Baby Hitler, etc.).