What, Errol Morris once asked Robert McNamara, protected humanity from extinction during the Cold War?1 Was it deterrence? Not at all. “We lucked out,” McNamara replied.2 Mankind had come within inches of the apocalypse, McNamara added, and it had come within inches of the apocalypse twenty or thirty times.
Little has changed. Cassandra has, if anything, grown more alarmed. The risk of nuclear catastrophe, former Secretary of Defense William Perry remarked, “is greater than it was during the Cold War and most people are blissfully unaware of this danger.”3 Perry made these remarks in 2016, well after the Cold War ended, and well before Donald Trump and Kim Jong-un began exchanging lurid threats.
For more than half a century, nuclear strategy has been governed by the doctrine of mutually assured destruction, or MAD; but nuclear strategy is, in part, the expression of an older view of international affairs, one in which every state faces a world in which there is a temptation to launch military action and a threat against doing so. In a world in which states are more or less equal, each state is apt to conclude that its best strategy is to do nothing. No state can do better, although every state can do worse. The balance of power worked well during parts of the nineteenth century, and especially between the Congress of Berlin in 1878 and the outbreak of the First World War in 1914. When it failed, it failed dramatically.
To traditional strategies, MAD adds a new concern—what Carl von Clausewitz called the escalation to extreme. Before 1945, states might have thought of war as offering advantages denied by peace. This is no longer true. Among nuclear states, escalation involves an outcome desired by no state and feared by all.
MAD is, by definition, a strategy of deterrence. If, as a strategy, it does not deter, it is no good. It is the promise of a second strike, if a first strike is launched, that comprises the deterrence. Yet an analysis of MAD suggests that, as a strategy, it is ineffective. Having been attacked by a first strike, every state understands that if its threat was a bluff, once called, it cannot be executed, and if was not a bluff, once challenged, it cannot have been a deterrent. If MAD fails as a deterrent, the strategy is ineffective. And worse. The conditions under which it might succeed as a threat guarantee that it must fail as a strategy. When Austria-Hungary served Serbia with an ultimatum in the summer of 1914, it threatened war if its ultimatum was rejected. Its ultimatum was rejected, and it went to war. It did not threaten worldwide annihilation, and it was unable to bring it about. This is the difference made by nuclear weapons. No state sees an advantage in a global nuclear catastrophe; it is not among any state’s goals. But if a second strike is not a rational goal, of what use is MAD as a rational strategy?
“Of no rational use at all,” is one obvious answer. A number of strategic theorists, and even a few politicians, have argued that MAD requires madmen in order to succeed. Who knows what they will do? This is an idea made plausible in Thomas Schelling’s The Strategy of Conflict; and made famous by Richard Nixon during the Vietnam War.4 Speaking to his chief of staff H. R. Haldeman, Nixon remarked:
I call it the Madman Theory, Bob. I want the North Vietnamese to believe I’ve reached the point where I might do anything to stop the war. We’ll just slip the word to them that, “For God’s sake, you know Nixon is obsessed about communism. We can’t restrain him when he’s angry—and he has his hand on the nuclear button,” and Ho Chi Minh himself will be in Paris in two days begging for peace.5
If MAD requires madmen to fail, it also requires madmen to succeed. When it comes to nuclear strategy, madmen are needed all around.
Fate and the Tiger
The unsettling declensions of modal paradox have persuaded some nuclear strategists that MAD does not require an endlessly receding hall of mirrors. No state need look into its neighbor’s eye. A deterrent intention is not necessary for deterrence to succeed. “[T]he existence of a nuclear retaliatory capability,” Gregory Kavka argued, “suffices for deterrence, regardless of a nation’s will, intentions, or pronouncements about nuclear weapons use.”6
“You don’t tangle with tigers,” David Lewis observes, “it’s that simple.”7
Tigers are themselves not needed. The human capacity for violence is tigerish enough to make the threat of retaliation plausible. According to René Girard, the human sense of the sacred arises through a similar mechanism. The sacred contains violence in the twofold sense of having something within itself and keeping it in check. One must not come too near to the sacred, for fear of provoking its violence, nor stand too far from it, for fear of losing its protection. “It is a curious paradox of our time,” Bernard Brodie has written, “that one of the foremost factors making deterrence really work and work well is the lurking fear that in some massive confrontation crisis it may fail.”8 Under these circumstances, he adds, one does not tempt fate. It is not clear that Brodie recognizes the terrible mixed metaphor in tempting fate. Nor is it clear that existential deterrence is an improvement on MAD. Both sides to a nuclear conflict may be deterred by nuclear tigers; one hopes that they are. But in order to be deterred by the nuclear tiger, each nuclear state must first believe that other states are in awe of its power.
Are they? It is easy enough to see that with this question, the remorseless logic of MAD reappears.
If nuclear strategy during the Cold War was governed by MAD, humanity’s survival was more miraculous than even McNamara supposed.
Toppling into Fatalism
Günther Anders was among the profound and radical thinkers to have reflected on the catastrophes of the twentieth century. Like Franz Kafka, he was inclined to parable. His retelling of the story of the flood is an example. In Anders’s version, Noah, clothed in sackcloth and ashes, tells the curious crowd that has gathered around him that he is in mourning for the victims of a flood that has not yet occurred. This could make sense only if the deaths to be mourned were inscribed in the future. Anders’s aim with this parable is to stress that the catastrophe, once it occurs, will have always been inevitable. The foretold suffering and deaths will occur. But no one is much motivated to avoid the inevitable. Nulla fata loco possis excludere, as Martial observed. What is needed is the delicacy of exhortation that allows Noah to say, “Let me help you build an ark, so that this may become false.”
The exhortation is moving because it is not clear that it is coherent. How can something that will always have been inevitable be avoided?
The world is what it is. No doubt. But whether it must be what it will be is another question. Determinism suggests that the future follows from the past and, expressed in this way, it seems unexceptional as a principle. From what else might it follow? The idea may be refined. Let the state of the world be represented as a maximally consistent set of sentences. Nothing true is left out. If the state of the world has been what it was, then every sentence in the set is true. Nothing false is let in. Among the sentences describing the state of the world are its laws of action. These permit predictions to be formulated as initial value problems, whether in physics or the theory of ordinary differential equations. Initial values, laws of nature, and predictions are all a part of the world’s state, the way that it is.
If these observations have an air of common sense, it is one which is very quickly dispelled. Determinism easily topples over into fatalism. A statement is necessary, logicians say, if it is true in every possible world, and possible, if it is not. It is an axiom of the modal system S5 that □(P → Q) → (□P → □Q). If □(P → Q) represents an inference to the future, Q is necessary if P is necessary. But P is a proposition that pertains to the past, and while many philosophers have rejected fatalism about the future, very few of them have scrupled at fatalism about the past.
The Moving Finger writes; and, having writ,
Moves on; nor all thy Piety nor Wit
Shall lure it back to cancel half a Line
Nor all thy Tears wash out a Word of it.9
If the future is determined, and the past is necessary, then the future must be necessary. What is this but fatalism? It is but a step from this conclusion to Diodorus. What is not necessary is impossible. What is not destined to take place cannot take place.
This is not an attitude one wishes to see encouraged among nuclear strategists. It leads again to paradox. Anyone who wishes to prevent a catastrophe must believe in its possibility. This Diodorus makes impossible. If a nuclear catastrophe is to occur, it must occur. Otherwise it is impossible. Deterrence is in the first case useless, and, in the second case, irrelevant.
An Appeal to a Higher Power
In some sense, fatalism about the future and fatalism about the past should be strictly symmetrical. Whatever is now in the past was once in the future, and whatever is now in the future will at some time be in the past. The symmetry is nonetheless imperfect. From the fact that the future is necessary it follows that it cannot be changed. Necessity comes first. From the fact that the past cannot be changed it follows that it is necessary. Unchangeability first. This asymmetry has given philosophers and analytic theologians room to maneuver. They have been surprisingly willing to maneuver backward by denying that all statements about the past are necessary. In this, they have intuition on their side. Trump might have lost the presidential election. In some possible world, he did.
This line of reasoning is not new; it has simply been reprised. Alvin Plantinga thus argues that what is unchangeable in the past is necessary, but he accepts the contrapositive without demur: what is not necessary in the past is not unchangeable. About necessity, he is resolute. Although one can discover that a sentence thought true is false, if a sentence is true, no agent can change what is true into what is false. This is itself, Plantinga argues, an accidentally necessary proposition. Such propositions have acquired their necessity in a moment of inadvertence. It is here that change and necessity diverge. That Trump won the presidential election is true in the real world—our own, as it happens, but it is a contingent proposition. Things could have been different. In some possible world, they were. But there is no possible world, at whatever modal distance from our own, such that in that world, it is false in our world that Trump won the presidential election.
What is at issue in this division of authority is the distinction between the proposition that P and the proposition that P is true-in-W. P may be false in some possible world, but in that world, that P is true-in-W remains true. “Indeed,” Plantinga writes, “for any proposition P and world W, if there is a world in which P-is-true in W, then P is true-in-W in every world.” True in every world, this proposition is necessary.
It is, one might say, just what it seems, and that is an accident of fate.
Like so many theologians, Plantinga has, in making this argument, appealed to a higher power—in this case higher-order modal logic. Contingent propositions about the past are about the world; the necessary propositions are about the contingent propositions. There is no single interpretive language that encompasses them both. The modal logician determined to talk about propositions requires a higher order of modal logic, and semantics on this level cannot be reduced downward to the language in which Trump may have failed to win the presidential election. This is the burden of Alfred Tarski’s great theorem about truth.10 Further concepts are required; they are required because richer concepts are required.
Still, Plantinga has offered an argument designed to cut off fatalism about the future by denying fatalism about the past. The modal axiom □(P → Q) → (□P → □Q) is fine. There is no need to retreat from S5. It is the initial premise in the argument that goes wrong. P is true, but not necessarily true.
Trump might have lost.
Friends of Freedom
If determinism is forever threatening to topple over into fatalism, then fatalism, once denied, can, in an exercise of contra-force majeure, easily come to threaten determinism. It is only some form of determinism, after all, that stands between a modal revision of certain propositions and a wholesale revision of the past. There is thus the imperative to reject the revision while retaining some possibility of change—just enough to make free will plausible. It is to this imperative that Lewis and Robert Stalnaker have responded.11 What Lewis calls “soft determinism” is
the doctrine that sometimes one freely does what one is predetermined to do; and that in such a case one is able to act otherwise though past history and the laws of nature determine that one will not act otherwise.12
The state of the world and its laws together predict what an agent will do. They themselves assign him no power to act otherwise. Those powers must be assumed. Their nature is not obvious. Lewis invites his reader to make a fine discrimination. The power to act otherwise comes in two forms, the strong and the weak. Granted the strong power, an agent is able to break the chains binding him to the past directly. This view demands of an agent the power either to violate the world’s laws of action or some among its necessary propositions. This is not plausible. The violation of determinism is too frank.
Under the weak version of Lewis-power, an agent is able to break his chains only if he is able to do something that if done would break them. This is counter-factual power. The agent remains a slave to the wheel of time. It is the wheel that changes. It changes counter-factually. But it is still a wheel, still in charge of things, and what an agent does is still determined. His power may be found, and must be exercised, against the grain, where the English modal system imposes a refined and subtle classification on the otherwise gross facts of life. Lewis thinks this good enough. The way in which an agent was determined not to do anything other than what he did, he writes, “was not the sort of way that counts as inability.”13
To the untrained eye, that is precisely what it seems—an inability.
Counter-Factual Control
No one would ever suggest that the free will championed by Lewis or Stalnaker is red blooded. It is the sort of thing that gets a man by and, as is so often the case, gets by him as well. Plantinga is as subtle as any man—hoch raffiniert, as German theologians would say—but he is more daring than most. A free agent, Plantinga argues, has the power to alter the past. This power cannot be causal. It is, nonetheless, real. It is the power to do something such that, if done, “the past will have been different from what it was in the actual world,” (emphasis added). This, Plantinga writes, is “counterfactual power over the past.”14 Whatever its effect on the past, counterfactual power requires the full resources of the English tense system simply to emerge into print.
Plantinga’s celebration of counterfactual power may suggest to some readers, at least, that the past admits of wholesale meddling. They are mistaken. The stress lies with “counterfactual”; and counterfactual power over the past does not involve changing the past. The idea that there may be more than one past is incoherent, and if there is only one past, nothing could be changed into anything. Counterfactual power over the past makes no such claims. The past is past. There is only one of it. But, subject to someone’s counterfactual control over the past, it would have been different. Although counterfactual power is power in this world, its influence is felt in some possible world. This is the meaning of the words “if he had done it,” which imply, among other things, that he has not done it. And it is a form of power over the past because if it had been exercised, then the past would necessarily have been other than what it was. Possibilities have come to influence possibilities.
There remains the question of just which propositions about the past are subject to a free agent’s counterfactual power. They are precisely, Plantinga argues, those propositions about the past that are not accidentally necessary. Trump might still have lost the election, but there is no counterfactual power over the necessary proposition that if “he won the election” is true in the real world, then it remains true in every world.
No power can touch that.
The Middle Passage
Counterfactual power over the past suggests a symmetrical power over the future. It is a power enjoyed and often exercised by the Christian God. Knowing what an agent would do allows God the luxury of considering those possible worlds in which what he would do follows from their prior state. God is then free to actualize the possible world that corresponds to what an agent would have done. Somehow an agent’s free will and God’s foreknowledge have both been accommodated. This is no mean trick. Whatever the trick, it is commonly assigned to the middle way of the Jesuit theologian Luis de Molina.
If counterfactual powers do not give an agent something that he cannot have (that is, the power to change the past), they are not nothing either. They are an interesting idealization of circumstances that occur often, and that occur in nuclear strategy. In many ways, nuclear strategy is an example of a pure promise, one unconstrained by contracts or their enforcement, and, in particular, nuclear strategy involves a promise contingent on a hypothetical. A promises to do something if B does something first. This is no different in logical scope from the comparable declaration made between nuclear states that if one of them strikes first, the other will strike second.
Why should we keep our promises, Heinrich Himmler asked in the closing days of World War Two, and the question provokes no very obvious answer. Why should we? Why should anyone? It is precisely in the case of these questions that counterfactual power, and its symmetrical counterfactual foreknowledge, make for something like an ideal point in complex analysis. In a world in which A keeps the promise made to B, B would have done something first. In a world in which A does not keep the promise made to B, B would not have done something first. If the promise is fulfilled, it is fulfilled in a possible world in which A fulfils his promise, and if not, then it takes place in a world in which he does not.
The power that Molina assigned to God involved both perfect counterfactual foreknowledge and a commensurate ability to bring certain worlds into being. In a world that is real and corrupt, agents tendering loans and states contemplating war lack divine counterfactual knowledge, but if they are free at all they are free to bring certain counterfactual states into being. A loan may be made or refused, a second strike launched or not, and each decision ushers in a different world.
The future must be such that the past it counterfactually determines does not causally prevent its own occurrence. This is no longer a possibility. Counterfactual power is not causal. If exercised to bring about a possible world fully commensurate with its counterfactual implications, that possible world, replete with its own deterministic structure, cannot bring about any world beyond that one to which it is now indissolubly linked. This is only true once the past has been determined, which, since the prior state of the world includes predictions about the future, it can only be once the future has been counterfactually determined.
This is the essential trait we have learned to ascribe to the metaphysics of the prophecy of doom.
The Residue of Uncertainty
In a world in which nuclear strategy were divinely controlled, or, at least, inspired, MAD would make sense. Each nuclear state makes a promise to respond to a first strike with a second strike. God has perfect foreknowledge of what his creatures will do. He may choose to bring into being a possible world consistent with the decision of every nuclear state faithfully to execute its dreadful counterfactual power. To make his strategy effective requires only that he spread the news.
And this he has not done.
What is left is the residue of uncertainty. Determinism—that remains. Fatalism—not so much. Both the escalation to the extreme and its negation are part of some future that is fixed, because both are the expression of some past over which agents have counterfactual power. If fatalism is flawed, nothing is forcing nuclear states to launch either a first or a second strike. It is because the former figures in the future that deterrence has a chance to work. It is because the latter figures in the future that the adversaries are not bound to destroy each other. Only the future will tell because only the future can tell.
This is, of course, horrifying.15
What remains unsaid in this discussion is a distinction unmade. The idea that the world is deterministic belongs to an old, magical way of thinking. It is our theories that are deterministic, or not. The world is what it is. Forces belong to an old physical tradition and are valuable only when there is nothing yet better to replace it. But neither do theories compel anything to occur in the world. How could they? Theories are so many words or symbols, and whatever else they may be doing, they are not in the business of compunction. The future is not determined by the past because neither the past nor the future is the sort of thing that determines anything. If deterministic theories provide no plausible account of free will, so much the worse for the theories.
And perhaps this, too, must be said and stressed.
Translated and adapted from the French by the editors.16
