The Worlds We Have Already Destroyed

Counterfactuals at the Heart of the Nuclear Paradox

Strategy, Geopolitics — Jul 9. Written by Gilles Deles-Velins.

On the occasion of the publication of our article “La honte de Prométhée” in Revue Projet — devoted to the parallel between the drone threat and that of nuclear weapons — we are attempting, in a series of posts, to unpack the concepts that surround this reflection.

Nuclear deterrence rests entirely on reasoning about events that do not occur. “If you attack us, we will retaliate.” That sentence, which sums up the logic of deterrence, is a conditional: it describes not what happens, but what would happen if things happened otherwise. Nuclear peace dwells, in its entirety, in a world that does not exist, and that must never exist. This paradox, far from being an academic curiosity, has occupied some of the most rigorous philosophers of the twentieth century, from David Lewis to Gregory Kavka, from David Gauthier to Jean-Pierre Dupuy. This article gives a rough sketch of the structure of their debate, starting from the conceptual tool that makes it possible: the counterfactual proposition.

* * *

I. What is a counterfactual?

The indicative and the counterfactual conditional

We all reason in counterfactuals, daily, without knowing it: “If I had left five minutes earlier, I could have caught my train”; “If I had revised more, I would have got a better grade.” Such sentences concern situations contrary to fact (I did not leave five minutes earlier, I did not revise more) and draw consequences from those situations about what would then have happened.

The distinction between the indicative conditional and the counterfactual conditional may seem trivial. In reality, it is momentous. Jean-Pierre Dupuy formulates it with remarkable clarity in The War That Must Not Occur. The proposition “If Shakespeare did not write Hamlet, someone else did” is indubitably true since the play exists and necessarily it has an author. By contrast, to assert the truth of the counterfactual proposition “If Shakespeare had not written Hamlet, someone else would have” is highly problematic; many may feel that only the Bard’s genius could have produced this masterpiece.¹ The shift in mood, from indicative preterite to past subjunctive, flips the reasoning into a different register. We are no longer reasoning about what is, but about what could have been. The truth of such propositions is no longer self-evident, yet their consequences in the Real are no less significant. The thought of the power of a “non-thing”² is dizzying and appears to us contradictory, and yet our world, its reliability, function continuously with these potential elements as their background.

David Lewis’s theory: possible worlds and similarity

It is David Lewis (1941–2001) who provided the most influential formalisation of counterfactuals. In Counterfactuals (1973), and then more extensively in “Counterfactual Dependence and Time’s Arrow” (1979), Lewis proposes what he calls Analysis 2: a counterfactual “If it were the case that A, then it would be the case that C” is (non-vacuously) true if and only if some (accessible) world in which both A and C are true is overall more similar to the actual world than is any world in which A is true but C is false.³

The key notion is that of overall similarity between possible worlds. Lewis readily concedes that this notion is vague, but he considers this vagueness a virtue rather than a defect: since counterfactuals are themselves both vague and various, an analysis that claimed to be fully precise would betray them.⁴

The temporal asymmetry of counterfactual dependence

But Lewis’s most decisive contribution for our purposes is his thesis on temporal asymmetry. The future depends counterfactually on the present, but the past generally does not. Lewis opens his 1979 article with a case of deceptive simplicity. He observes that he is typing words on a page, and invites us to suppose that he were typing different ones. Tomorrow would obviously be different — different words would appear on the page. But would yesterday be different too? And if so, how? One can no doubt find something to say, Lewis concedes, but nothing that would seem clearly and uncontroversially true.⁵

Lewis holds that this asymmetry is a fundamental feature of our ordinary counterfactual reasoning: if the present were different, the future would be different, but the past would remain the same. Its trace can be found in the most banal reasoning. If I were acting otherwise just now, I would revenge a wrong done me last year — it is absurd even to raise the question whether that past wrong would have taken place if I were acting otherwise now.⁶ This point, apparently abstract, is the keystone of the entire debate on nuclear deterrence. For it is precisely this asymmetry (the past is fixed, the future is open) that makes the paradoxes of deterrence so formidable.

* * *

II. When counterfactuals meet the bomb

The logical structure of deterrence

To deter is to render credible a counterfactual conditional: “If you attack us, we will retaliate.” But deterrence has a peculiar feature: if it works, the conditional is never tested. The attack does not occur, nor does the retaliation. The proposition remains suspended in a world that never comes to pass.

It was Gregory Kavka who first formalised the moral paradoxes this structure generates, in a founding 1978 article. Kavka there defines Special Deterrent Situations (SDSs): an agent finds himself in such a situation when he reasonably and correctly believes that he must form the conditional intention to apply a harmful sanction to innocent people in order to prevent a very harmful and unjust offence, that such an intention would very likely deter the offence, and that he would have conclusive moral reasons not to apply the sanction if the offence were to occur.⁷

Kavka’s introductory example is directly that of nuclear deterrence. A nation N sees the threat of nuclear retaliation as its only reliable means of preventing an attack or blackmail by its rival. N is confident that the threat will succeed in deterring its adversary, provided that it really intends to carry it out. But N also recognises that it would have conclusive moral reasons not to carry out the threatened retaliation if its opponent were to obliterate N with a surprise attack, for retaliation would destroy many millions of innocent civilians and might add enough fallout to the atmosphere to destroy the human race.⁸

Kavka’s three paradoxes

From this situation, Kavka draws three paradoxes.

The first states that there are cases in which, although it would be wrong for an agent to perform a certain act in a certain situation, it would nonetheless be right for him, knowing this, to form the intention to perform that act in that situation.⁹ This flatly contradicts what Kavka calls the Wrongful Intentions Principle (WIP), which may be formulated thus: to intend to do what one knows to be wrong is itself wrong. The WIP is so universally accepted that philosophers scarcely ever take the trouble to spell it out. Kavka notes that it has been accepted, at least implicitly, by Abelard, Aquinas, Butler, Bentham, Kant, and Sidgwick.¹⁰ And yet nuclear deterrence seems to invalidate it, for the formation of the intention (threatening massive retaliation) produces considerable autonomous effects, independent of the act itself: what prevents catastrophe is the intention, not the act. The moral evaluation of the intention cannot therefore be reduced to the moral evaluation of the act it aims at, since the peculiar property of a deterrent intention is to be self-defeating: if it fulfils its function, it guarantees precisely that the (immoral) act will never occur.¹¹

The second paradox is more cruel still. Kavka shows that an agent who is at once rational and morally good finds himself in a state of logical impossibility as regards forming the retaliatory intention that the first paradox says he ought to form. There are situations (namely SDSs) in which it would be right for agents to perform certain actions (namely forming the intention to apply the sanction) and in which it is possible for some agents to perform such actions, but impossible for rational and morally good agents to perform them.¹²

The rational and moral agent cannot dispose himself to do what he knows he would have conclusive reasons not to do. Rationality and virtue, far from making just action easier, make its accomplishment impossible. Kavka puts it with a striking image: the moral agent is a captive in the prison of his own virtue, able to form the requisite intention only by bending the bars of his cell out of shape.¹³

The toxin puzzle: the parable

To give a non-nuclear version of the same paradox, Kavka invented what has come to be called the “toxin puzzle.”¹⁴ Dupuy offers a particularly eloquent retelling. An eccentric billionaire offers you a deal: if, tonight at midnight, you sincerely form the intention of drinking tomorrow at noon a vial of toxin that will make you sick for two days, he will deposit a million dollars in your account tomorrow morning. He does not even ask you to drink the toxin; it suffices that you form the intention, detected by a machine of his own invention.

Yet, as Kavka shows, a rational agent finds himself unable to form this intention. For tomorrow at noon, whether the million has been paid or not, the past is fixed. Drinking has no utility and causes harm. Not drinking is the dominant strategy, and the agent knows this from the start. As Dupuy writes: because forming an intention is not a volitional act, it is impossible for you to form the intention to do x if you know that, when the moment comes, it will be unreasonable for you to do x.¹⁵

This is exactly the structure of the nuclear paradox: it is rational to threaten retaliation, but a fully rational agent cannot sincerely form the intention to carry it out, because he knows that when the moment comes, carrying it out would be irrational and immoral. Dupuy condenses the impasse in a striking formula: nuclear deterrence creates a situation in which a rational subject cannot do the very thing that it is rational for him to do.¹⁶ Thus, the entire structure of deterrence holds together thanks to a performative and rhetorical artifice in which it is important, like Nixon, to play the madman. The problem is that in playing the madman one may become the character one enacts and impersonates, pointing to the danger that lies in the shadows, and this danger can only be neutralised by a greater danger. Are inter-state politics then condemned to a power struggle? David Lewis suggests a possible way out by problematising things from a different slope.

* * *

III. David Lewis facing the paradox: deterrent existentialism

The essential philosopher

David Lewis intervenes in the deterrence debate at several levels, which makes him an essential author. His theory of counterfactuals provides the logical grammar in which the problem is formulated. His work on temporal asymmetry provides the premise that makes the toxin puzzle insoluble: since the past is counterfactually fixed, tomorrow’s decision cannot “reach back in time” to affect today’s payment. And his original strategic position, developed in “Finite Counterforce,” proposes a dissolution of the paradox through a bold empirical thesis: deterrent existentialism.

The dissolution through existentialism

Lewis begins by unfolding the inexorable logic that leads either to the “MAD” position (Mutual Assured Destruction: if we cannot be credible, we must be terrifying), or to the “NUT” position (Nuclear Utilization Theorist: if we cannot be terrifying, we must be credible). The starting point is simple: we deter the enemy from doing x by threatening that if he does, then we will punish him by doing y. But the enemy might notice that if he does x, we will then have no good reason to do y. We therefore have a credibility problem: our deterrence is apt to fail because our threats are not believed.¹⁷

Lewis’s solution runs through what he calls “existentialism,” an empirical thesis about the psychology of the enemy. The enemy is sceptical, pessimistic, risk-averse, and conservative. He is not a confident calculator who precisely measures the probability of retaliation: he is an anxious actor who fears the worst.¹⁸ Under these conditions, deterrence functions not through the credibility of the declaratory threat, but through the mere existence of destructive capacities. Lewis puts it tersely: we deter by what we have, not by what we say. If we have the capacity to destroy cities, and we have evident reasons not to, and we say we will, the pessimistic and sceptical enemy will think we maybe will and maybe won’t, and he will fear the worst. If we say we won’t, but we keep the capacity, then too he will think we maybe will and maybe won’t.¹⁹

Lewis draws from this an original practical position, which he sums up in the formula “buy like a MADman, use like a NUT”: buy a modest arsenal, sufficient for its very existence to deter, but without needing to seek to render credible our dreadful threat to destroy the enemy’s cities. We need not even utter that threat. We have weapons that give us the assured capacity to do it, and their very existence is deterrent enough.²⁰

Dodging the paradox

If Lewis’s existentialism is true, Kavka’s “paradox of deterrence” simply does not arise. Lewis is explicit: if existentialism is true, and it is not intentions but capacities that deter, then deterrence does not require us to cultivate intentions to do anything except what would at the time be right. The “paradox of deterrence,” in which supposedly it is intentions that deter, does not arise.²¹

The move is elegant: rather than resolving the moral paradox, Lewis contests its empirical premise. If it is capacities and not intentions that deter, then there is no need to form the intention to do anything immoral, and the conflict between rationality and morality melts away.

Lewis had expressed this position more polemically in an earlier text, “Devil’s Bargains and the Real World,” in which he reproached those who took the paradox seriously with drawing a picture that, in his view, implicitly slanders many decent patriots in the American armed forces and in the White House.²² Fred Kroon, commenting on this position, notes that Lewis considers actual nuclear deterrers to be a “strange” mixture of good and evil, and of the rational and the irrational, while conceding in conversation that he is a clear case of an agent-irrationalist only if the standards for full rationality are set very — perhaps inappropriately — high.²³

* * *

IV. Jean-Pierre Dupuy: the counterfactual reversed

Occurring time and projected time

Dupuy takes up the problem where Lewis leaves it, but reverses its temporal logic. His starting point is that classical strategic thought, including Lewis’s, operates in what he calls “occurring time.” In this metaphysics, the past is fixed and the future is open; the agent chooses among possible branches by taking the past to be counterfactually independent of his decision. This is the metaphysics of the decision tree, the metaphysics of game theory, the metaphysics of all standard strategic rationality. Dupuy characterises it thus: at every node of the tree an agent has a choice between several possible future courses of action; when he chooses among them, he takes the past to be fixed, which is to say counterfactually — and not only causally — independent of his choosing. Fixed past and open future define this temporality.²⁴

Dupuy proposes an alternative metaphysics that he names “projected time.” In it, the future is fixed (necessary once actualised), and it is the past that becomes counterfactually dependent on the future action. The future retroactively determines the past conditions that made it possible. The indeterminacy of the past so long as an action has not taken place, together with the necessity of the future once an action has taken place, define this metaphysics of temporality.²⁵ The future is then the “fixed point” of an operator F which expresses the causal consequences of a past that is itself determined counterfactually by the future action. The formula is x = F[x]: the action is a function of itself, as if it were determining itself.²⁶

Diodorus Cronus and the plurality of metaphysics

To justify the coexistence of these two metaphysics, Dupuy goes back to the aporia of Diodorus Cronus (fourth century BCE), who had shown the incompatibility of three axioms: (1) Every true proposition about the past is necessarily true; (2) The impossible cannot be deduced from the possible; (3) Something that neither is nor will be is possible. Occurring time retains axioms 1 and 3 but rejects 2; projected time rejects axioms 1 and 3.²⁷ The two metaphysics are therefore two coherent ways of resolving the same aporia, and, as Dupuy stresses when quoting Poincaré on non-Euclidean geometries, it is pointless to ask whether the metaphysics of projected time is truer than that of occurring time: the real question is whether it is more convenient.²⁸

Deterrence in projected time

Applied to nuclear deterrence, the metaphysics of projected time first produces a negative result. If the future is necessary, no form of deterrence works. For if deterrence succeeds, the event it is meant to prevent (escalation to extremes) does not occur. In projected time, where the possible does not exceed the actual, that event is therefore impossible. But if it is impossible, the threat that mobilises it has no reality, and deterrence collapses. Dupuy formalises this reductio in five propositions and concludes: if deterrence is effective, then it is not effective.²⁹

The detour seems unproductive. But Dupuy then introduces a concept that changes everything: the indeterminacy of the future in projected time. Unlike the probabilistic uncertainty of occurring time (in which subjective probabilities are assigned to disjunctive scenarios), indeterminacy in projected time takes the form of a superposition of states: escalation to extremes and peace both figure in the fixed future, not as alternative possibilities but as superposed states. The conclusion Dupuy reaches is that what has allowed nuclear deterrence to work until now, and may allow it to go on working, is the indeterminacy of the future in a conception of time that makes the future necessary.³⁰

Catastrophe is “inevitable and exact, but it may not happen,” in the Borgesian formula Dupuy takes up as his own.³¹ This is the programme of what he once called “enlightened doomsaying”: neither fatalism (catastrophe is certain, let us give up) nor smug optimism (the future is open, nothing compels us to act). By granting the agent the retrograde counterfactual power to act on the past conditions that push him to act, projected time makes it possible to steer between the two pitfalls.³² The idea of a reversal of the arrow of time, in which a future state retroactively determines the conditions of its own coming to pass, is not, moreover, without kinship with other theoretical intuitions. Henri Atlan had already proposed, in Entre le cristal et la fumée, to think of DNA as a programme for the elaboration of a biological subject yet to come — that is, as a form of determination of the past by the future — which grants Dupuy’s temporal reversal an anchoring that exceeds purely speculative metaphysics.³³

The figure of the prophet

This temporal reversal is not, for Dupuy, an abstract invention. He finds its trace in the figure of the biblical prophet, the nabi, who must solve a fixed-point problem: to predict a future that the people will take to be an unalterable given, while knowing that this future will be caused in part by the prophecy he publicly makes of it. From this, Dupuy writes, there results a highly paradoxical combination of fatalism and voluntarism: fatalism on the side of the people, who take the word of the prophet to be the word of God, and voluntarism (at least partially) on the side of the prophet, who knows that his words have a causal impact on the future.³⁴

* * *

V. What the debate reveals

The heart of the problem is that nuclear deterrence is a counterfactual machine: it works only in a world that never comes to pass, the one in which the attack takes place and the retaliation falls. Its entire rationality rests on reasoning about situations that are never realised — failing which it fails.

If the only weapon that protects us functions entirely in an imaginary world, a world in which we would have pulled the trigger, what do we truly know about the solidity of that protection? This is a question no experiment can settle for us.

Notes

1. Jean-Pierre Dupuy, The War That Must Not Occur, translated by Malcolm B. DeBevoise, Stanford, Stanford University Press, 2023, note 30 [French original: La guerre qui ne peut pas avoir lieu. Essai de métaphysique nucléaire, Paris, Desclée de Brouwer, 2018, note 37]. The complete original of the note: “The linguistic and metaphysical tool that permits a comparison to be made between the actual world and another possible world is called a counterfactual proposition. […] The term counterfactual refers to the presence of an antecedent […] that is contrary to fact.” ↩

2. The notion of a “non-thing” and the causal power of what does not exist would invite a parallel with the “work of the negative” in André Green, for whom the negative is not the absence of something but a structuring force in its own right. Establishing that kinship rigorously would however require an article of its own. See André Green, The Work of the Negative, translated by Andrew Weller, London, Free Association Books, 1999. ↩

3. David Kellogg Lewis (1941–2001), American philosopher, professor at Princeton University, is one of the major figures of contemporary analytic philosophy. See his Wikipedia page. ↩

4. Lewis, “Counterfactual Dependence and Time’s Arrow,” Noûs, vol. 13, no. 4, November 1979, p. 465. In the original: “This analysis is fully general: A can be a supposition of any sort. It is also extremely vague. […] I count that a virtue. Counterfactuals are both vague and various.” ↩

5. Lewis, “Counterfactual Dependence and Time’s Arrow,” art. cit., p. 455 (incipit of the article). In the original: “Today I am typing words on a page. Suppose today were different. Suppose I were typing different words. Then plainly tomorrow would be different also; for instance, different words would appear on the page. Would yesterday also be different? If so, how? Invited to answer, you will perhaps come up with something. But I do not think there is anything you can say about how yesterday would be that will seem clearly and uncontroversially true.” ↩

6. Lewis, ibid., p. 456. In the original: “If I were acting otherwise just now, I would revenge a wrong done me last year — it is absurd even to raise the question whether that past wrong would have taken place if I were acting otherwise now!” ↩

7. Gregory S. Kavka, “Some Paradoxes of Deterrence,” The Journal of Philosophy, vol. 75, no. 6, June 1978, p. 286–287. JSTOR: 2025707. In the original, Kavka defines an SDS as a situation in which an agent “reasonably and correctly believes” that the following conditions hold: “First, it is likely he must intend (conditionally) to apply a harmful sanction to innocent people […]. Finally, he would have conclusive moral reasons not to apply the sanction if the offense were to occur.” ↩

8. Kavka, “Some Paradoxes of Deterrence,” art. cit., p. 286. In the original: “N recognizes it would have conclusive moral reasons not to carry out the threatened retaliation, if its opponent were to obliterate N with a surprise attack. […] N knows it would also destroy many millions of innocent civilians […] and might add enough fallout to the atmosphere to destroy the human race.” ↩

9. Kavka, ibid., p. 288. In the original: “(P1) There are cases in which, although it would be wrong for an agent to perform a certain act in a certain situation, it would nonetheless be right for him, knowing this, to form the intention to perform that act in that situation.” ↩

10. Kavka, ibid., p. 289. In the original: “To intend to do what one knows to be wrong is itself wrong.” Kavka notes that this principle has been accepted by “Abelard, Aquinas, Butler, Bentham, Kant, and Sidgwick.” ↩

11. Kavka, ibid., p. 291. In the original: “WIP ties the morality of an intention exclusively to the moral qualities of its object (i.e., the intended act). […] However, in certain cases, intentions may have autonomous effects that are independent of the intended act’s actually being performed. In particular, intentions to act may influence the conduct of other agents. When an intention has important autonomous effects, these effects must be incorporated into any adequate moral analysis of it.” ↩

12. Kavka, ibid., p. 294. Formulation of the second paradox (P2). In the original: “(P2) There are situations (namely SDSs) in which it would be right for agents to perform certain actions (namely forming the intention to apply the sanction) and in which it is possible for some agents to perform such actions, but impossible for rational and morally good agents to perform them.” ↩

13. Kavka, ibid., p. 291. In the original: “He is a captive in the prison of his own virtue, able to form the requisite intention only by bending the bars of his cell out of shape.” ↩

14. Kavka develops the toxin puzzle in “The Toxin Puzzle,” Analysis, vol. 43, no. 1, January 1983, p. 33–36. The parable also appears in Gregory S. Kavka, Moral Paradoxes of Nuclear Deterrence, Cambridge, Cambridge University Press, 1987, p. 47. ↩

15. Dupuy, The War That Must Not Occur, op. cit., chap. 3, section on the toxin puzzle. In the original: “Because forming an intention is not a volitional act, it is impossible for you to form the intention to do x if you know that, when the moment comes, it will be unreasonable for you to do x.” ↩

16. Dupuy, ibid., same section. In the original: “Nuclear deterrence creates a situation in which a rational subject cannot do the very thing that it is rational for him to do.” ↩

17. David Lewis, “Finite Counterforce,” in Henry Shue (ed.), Nuclear Deterrence and Moral Restraint, Cambridge, Cambridge University Press, 1989, p. 53. DOI: 10.1017/CBO9780511625145.005. In the original: “We deter the enemy from doing x by threatening that if he does, then we will punish him by doing y […]. But the enemy might notice that if he does x, we will then have no good reason to do y. […] In short, we have a credibility problem: Our deterrence is apt to fail because our threats are not believed.” ↩

18. Lewis develops the thesis of existentialism in pages 51–73 of “Finite Counterforce.” He borrows the term “existential deterrence” from McGeorge Bundy. His formulation: “My task will not be to argue for existentialism, but to argue from it” (p. 53). ↩

19. Lewis, “Finite Counterforce,” art. cit., p. 76. In the original: “We deter by what we have, not by what we say. If we have the capacity to destroy cities, and we have evident reasons not to, and we say we will, the pessimistic and skeptical enemy will think we maybe will and maybe won’t, and he will fear the worst. If we say we won’t, but we keep the capacity, then too he will think we maybe will and maybe won’t.” ↩

20. Lewis, ibid., p. 73. In the original: “We can buy like a MADman, if we like, but that implies nothing about what we ought to do in case of war, or what we ought to intend beforehand. We needn’t strive to give some credibility to our dreadful threat to destroy the enemy’s cities. We needn’t threaten it at all. We have weapons that give us the assured capacity to do it, and their very existence is deterrent enough.” ↩

21. Lewis, ibid., p. 74. In the original: “If existentialism is true, and it is not intentions but capacities that deter, then deterrence does not require us to cultivate intentions to do anything except what would at the time be right. The ‘paradox of deterrence’, in which supposedly it is intentions that deter […] does not arise.” Lewis refers in a note to Kavka, “Some Paradoxes of Deterrence,” and to Gauthier, “Deterrence, Maximization, and Rationality.” ↩

22. David Lewis, “Devil’s Bargains and the Real World,” in Douglas MacLean (ed.), The Security Gamble: Deterrence Dilemmas in the Nuclear Age, Totowa, N.J., Rowman & Allanheld, 1984, p. 148. In the original: “a picture of nuclear deterrence that implicitly slanders many decent patriots in the American armed forces and in the White House.” ↩

23. Fred Kroon, “Deterrence and the Fragility of Rationality,” The Monist, vol. 79, no. 3, 1996, p. 354–355. In the original: “Lewis insists that our (hypothetical) nuclear deterrers are a ‘strange’ mixture of good and evil, and of the rational and the irrational. […] In conversation, however, Lewis claims that he is a clear case of an agent-irrationalist only if the standards for full rationality are set very — perhaps inappropriately — high.” ↩

24. Dupuy, The War That Must Not Occur, op. cit., Chapter 4 “Metaphysical MAD,” section “The Force of Destiny.” In the original: “At every node of the tree an agent has a choice between several possible future courses of action. When he chooses among them, he takes the past to be fixed, which is to say counterfactually — and not only causally — independent of his choosing.” ↩

25. Dupuy, ibid. In the original: “The indeterminacy of the past so long as an action has not taken place, together with the necessity of the future once an action has taken place, define a metaphysics of temporality that I call ‘projected time.’” ↩

26. Dupuy, ibid. In the original: “Following the conventional terminology, let us say that the future is the fixed point of a certain operator F. This operator expresses the causal consequences of a past that itself is determined counterfactually by a future action x.” ↩

27. Dupuy, ibid. Dupuy attributes the formalisation of Diodorus’s incompatibility theorem to the philosopher Jules Vuillemin. ↩

28. Dupuy, ibid., quoting Henri Poincaré. In the original: “One geometry cannot be more true than another; it can only be more convenient.” ↩

29. Dupuy, ibid., section “Nuclear Deterrence in Projected Time.” The reductio is developed in five numbered propositions, the conclusion of which is: “It follows from this that if nuclear deterrence works, then it does not work. Therefore nuclear deterrence does not work.” ↩

30. Dupuy, ibid., section “Nuclear Deterrence and the Indeterminacy of the Future.” The central thesis of the work, italicised by the author. In the original: “what has allowed nuclear deterrence to work until now, and may allow it to go on working, is the indeterminacy of the future in a conception of time that makes the future necessary.” ↩

31. Dupuy, ibid. In the original: “The future is inevitable and exact, but it may not happen. God lies in wait in the intervals.” The formula is taken from Jorge Luis Borges, “The Creation and P. H. Gosse,” in Selected Non-Fictions, ed. and trans. Eliot Weinberger, New York, Penguin, 1999, p. 223. The thesis is developed in Jean-Pierre Dupuy, How to Think About Catastrophe: Toward a Theory of Enlightened Doomsaying, translated by M. B. DeBevoise and Mark R. Anspach, East Lansing, Michigan State University Press, 2022 [French original: Pour un catastrophisme éclairé. Quand l’impossible est certain, Paris, Seuil, 2002]. ↩

32. Dupuy, ibid., section “The Force of Destiny.” In the original: “By granting agents the counterfactual power to act on the past, which is to say on the conditions that determine their behavior, projected time helps them steer a safe course between the devil of doomsaying and the deep blue sea of smug optimism.” ↩

33. Henri Atlan, Entre le cristal et la fumée. Essai sur l’organisation du vivant, Paris, Seuil, 1979 [not translated into English]. Atlan there develops the idea that the genetic programme can be understood as a retrograde determination: it is the future organism that retrospectively confers its meaning on the DNA, rather than the DNA that “contains” mechanically the organism to come. ↩

34. Dupuy, The War That Must Not Occur, op. cit., Chapter 4, section “The Force of Destiny.” In the original: “From this, there results a highly paradoxical combination of fatalism and voluntarism: fatalism on the side of the people, who take the word of the prophet to be the word of God, and voluntarism (at least partially) on the side of the prophet, who knows that his words have a causal impact on the future.” ↩

Principal references

• Atlan, Henri, Entre le cristal et la fumée. Essai sur l’organisation du vivant, Paris, Seuil, 1979 [not translated into English].
• Dupuy, Jean-Pierre, The War That Must Not Occur, translated by Malcolm B. DeBevoise, Stanford, Stanford University Press, 2023 [French original: La guerre qui ne peut pas avoir lieu. Essai de métaphysique nucléaire, Paris, Desclée de Brouwer, 2018].
• Dupuy, Jean-Pierre, How to Think About Catastrophe: Toward a Theory of Enlightened Doomsaying, translated by M. B. DeBevoise and Mark R. Anspach, East Lansing, Michigan State University Press, 2022 [French original: Pour un catastrophisme éclairé. Quand l’impossible est certain, Paris, Seuil, 2002].
• Gauthier, David, “Deterrence, Maximization, and Rationality,” Ethics, vol. 94, no. 3, April 1984, p. 474–495.
• Kavka, Gregory S., “Some Paradoxes of Deterrence,” The Journal of Philosophy, vol. 75, no. 6, June 1978, p. 285–302.
• Kavka, Gregory S., Moral Paradoxes of Nuclear Deterrence, Cambridge, Cambridge University Press, 1987.
• Kroon, Fred, “Deterrence and the Fragility of Rationality,” The Monist, vol. 79, no. 3, 1996, p. 349–370.
• Lewis, David, Counterfactuals, Oxford, Blackwell, 1973.
• Lewis, David, “Counterfactual Dependence and Time’s Arrow,” Noûs, vol. 13, no. 4, November 1979, p. 455–476.
• Lewis, David, “Devil’s Bargains and the Real World,” in Douglas MacLean (ed.), The Security Gamble, Totowa, N.J., Rowman & Allanheld, 1984, p. 141–154.
• Lewis, David, “Finite Counterforce,” in Henry Shue (ed.), Nuclear Deterrence and Moral Restraint, Cambridge, Cambridge University Press, 1989, p. 51–114.