Free Will

Peter Van Inwagen presents an argument suggesting that free will, if understood as requiring that we could act otherwise than we do, is incompatible with determinism. This is a very significant claim, since if it were true, the only option for compatibilists would be to deny the widely-accepted thesis that freedom requires such an ability.[1] Many compatibilists have favoured a hypothetical analysis of ‘S could have done otherwise’, interpreting the phrase as meaning something along the lines of ‘S would have done otherwise if he had wanted to’. However, if Van Inwagen is correct that any genuine analysis of ‘could have done otherwise’ is incompatible with determinism, then the compatibilist tradition of the hypothetical analysis would be shown to be mistaken. It is the purpose of this essay to examine Van Inwagen’s argument, and assess whether his claims are justified.

In a sense, the argument is actually rather superfluous, for its conclusion will either be plainly true or plainly false, depending on our interpretation of the word ‘could’. If future states of affairs are precisely determined by past ones, then we obviously cannot (categorically) bring about any state of affairs other than that one which is necessitated. It would hardly be worth writing up a complex argument for this undisputed triviality.

Alternatively, if we adopt the hypothetical analysis of ‘could’, then Van Inwagen’s conclusion is simply, undeniably, false. It is entirely consistent with determinism to recognise that if, contrary to fact, a different past state of affairs had obtained, then, via the deterministic laws of nature, a different current state of affairs would inevitably result. But to say ‘S would have done otherwise if he had wanted to’ is just to say that if, contrary to fact, S’s past desires had been different, then S’s current behaviour would likewise differ. Thus the hypothetical analysis is undeniably consistent with determinism, and any proposition to the contrary is undeniably false. Nevertheless, the premises of Van Inwagen’s argument all have a certain prima facie plausibility. Since the argument is logically valid, the only way the conclusion could be false is if it contains a false premise. So it remains a worthwhile – and far from trivial – exercise, to identify the fallacious premise(s).

We also face a further challenge of far greater import. Van Inwagen concedes that the hypothetical analysis may entail the falsity of some premise of his argument.[2] He then suggests that, because his premises are all true, the ‘falsified’ premise would merely stand as a counterexample to the hypothetical analysis. In other words, Van Inwagen is arguing that because his premises are all more plausible than the hypothetical analysis of the word ‘could’, if they come into conflict, we should conclude that the hypothetical analysis must have been mistaken. Thus, if we are to find a flaw in his main argument, it is not enough to merely show a premise to be false according to the hypothetical analysis of ‘could’. We must go further, and give a convincing general explanation which dispels the initial plausibility of the premises. If their certainty can be cast into doubt, they will no longer serve as counterexamples to the hypothetical analysis, so Van Inwagen’s claims would prove to be unfounded.

Van Inwagen’s ‘main argument’ follows:

I shall use ‘T0’ to denote some instant of time earlier than J’s birth, ‘P0’ to denote the proposition that expresses the state of the world at T0, ‘P’ to denote the proposition that expresses the state of the world at T, and ‘L’ to denote the conjunction into a single proposition of all laws of physics…

1) If determinism is true, then the conjunction of P0 and L entails P.

2) If J had raised his hand at T, then P would be false.

3) If (2) is true, then if J could have raised his hand at T, J could have rendered P false.

4) If J could have rendered P false, and if the conjunction of P0 and L entails P, then J could have rendered the conjunction of P0 and L false.

5) If J could have rendered the conjunction of P0 and L false, then J could have rendered L false.

6) J could not have rendered L false.

Therefore, 7) If determinism is true, J could not have raised his hand at T.[3]

To summarise the argument in more general terms: ‘J could have done otherwise’ implies that ‘J could have rendered the current state of affairs (P) false’. But according to determinism: the past state of the universe (P0), in conjunction with the laws of nature (L), together necessitate the present state of affairs (P). So to render P false, you must render false either P0 or L. But surely J could not change the state of the universe before he was born, nor break the laws of nature. So J could not render P false. Therefore, determinism implies that J could not have done other than he did.

As an initial observation, I want to highlight the fact that according to the most natural interpretation of ‘render’ (as meaning something along the lines of ‘make’ or ‘cause to be’),[4] premise #4 is quite definitely mistaken. It rests upon the patently false assumption that to cause something, you must be the ultimate cause of it. For if you are in the middle of the causal chain, you will cause (‘render’) the later events, but not the earlier ones. Thus we have ourselves a simple counterexample: Suppose the conjunction of P0 and L is false independently of J’s actions (i.e. it is false, though J did not render it false). It is nevertheless entirely possible for J’s actions to causally contribute to the falsity of P (i.e. render P false). Indeed, this is precisely what a determinist would expect to happen: different past conditions would cause J to have different desires, which in turn cause J to act differently.[5] So the causal chain ~(P0 & L)-> J -> ~P stands as a counterexample to the natural interpretation of premise #4, because here J can render P false, though he cannot render false the conjunction of P0 and L (because it was already false).

Van Inwagen, however, intends his use of the word ‘render’ to be understood differently. According to his definition, someone can render a proposition false if it is “within his power to produce any set of conditions sufficient for the falsity of this proposition”.[6] This does not require any connection between the actor and the proposition’s contents, beyond a merely logical one. This leads to some highly counterintuitive results. Particularly, it means that for any false proposition P, anyone can ‘render P false’ by doing anything at all.[7] For example, he would say that I can render 1 + 1 = 3 false, just by blinking, and by clapping my hands I can render it false that Al Gore is president of the USA. So we should recognise that Van Inwagen is using the word ‘render’ in a non-standard way. This in itself is not necessarily a problem,[8] but it does mean that Van Inwagen has forfeited the right to appeal to our intuitions regarding the word ‘render’ in defence of his later premises. Taking ‘render’ as Van Inwagen intended, then, results in premise #4 being an analytic truth,[9] as unassailable as the first three premises. Let us move on to considering the next premise instead.

Van Inwagen takes it as analytic that S cannot render false any “true proposition that concerns only states of affairs that obtained before S’s birth”.[10] The initial plausibility of this depends upon our intuition that we cannot change the past. Using the natural interpretation of ‘render’, it strikes us as obvious that we cannot presently ‘render false’ any proposition wholly about the past. However, recall that Van Inwagen is using the term in a non-standard way. In particular, note that all it would take for us to ‘render false’ a past proposition, would be for that proposition to be false. So this premise (#5) is not really about whether we can render false past propositions (since we obviously can), but rather, whether we can render false true propositions.

If ‘can’ is understood categorically, then Van Inwagen may well be correct in suggesting that it is analytic that we cannot ‘render false’ true propositions wholly concerning the past. For, if understood categorically, it seems analytic that we cannot render false any true proposition at all.[11] Suppose the proposition Q: ‘Tomorrow, S will perform act A’ is true. Then we cannot (categorically) do anything which would entail the falsity of Q, because it is not false, and we surely cannot (categorically) bring about a logically inconsistent state of affairs.

This does provide some reason to doubt the usefulness of categorical interpretations of ‘could’. But my main point here is that the special distinction Van Inwagen wants to draw between past and future propositions cannot be achieved on purely logical grounds. In rejecting the natural interpretation of ‘render’ as he does, Van Inwagen has lost the ability to justify this distinction, as it rests so heavily on the asymmetrical directionality of causal notions;[12] the very notions he must ignore in order to preserve premise #4. So it seems that if he is to be consistent, Van Inwagen is committed to a form of fatalism – a commitment he would very much want to avoid.[13]

But just as fatalist arguments can be undone by noting the conceptual distinction between future event and inevitable future event,[14] so Van Inwagen’s defence of premise #5 can be undone by distinguishing truth from necessary truth.[15] It is entirely possible that the past could have been different from how it was. The proposition P0, though true, could have been false. Thus, it is entirely reasonable to say that J could (understood hypothetically, of course) render P0 false. I should emphasise that this is not the preposterous claim that J could change the past. Rather, I am observing the undeniable fact that the past could have been different, which would be a sufficient condition for J’s ‘rendering’ P0 false, according to Van Inwagen’s definition. This seems to be a perfectly legitimate use of the word ‘could’, which implies the falsity of premise #5.[16] At the very least, I have shown that the premise is not so certain as to disprove the compatibilist’s hypothetical analysis of ‘could’.

Ultimately, Van Inwagen’s argument for the incompatibility of free will and determinism, is little more than rhetorical window-dressing for his ‘categorical’-leaning intuitions regarding the analysis of ‘could’. He misleadingly appropriates the word ‘render’ to bolster the intuitive plausibility of his later premises. But once interpreted in terms of his non-standard use of the word, they become much less convincing. This casts significant doubt on Van Inwagen’s assertion that the inconsistency between his premises and the hypothetical analysis of ‘could’ demonstrates the falsity of latter. The argument is unsound according to the compatibilist’s hypothetical analysis, and it offers precious little reason to suggest that this analysis is wrong. The gulf which separates the intuitions of compatibilists from incompatibilists has proven too wide for Van Inwagen’s argument to cross.

Bibliography

Dennett, D. ‘I Could Not Have Done Otherwise – So What?’ in R. Kane (ed.) Free Will, Oxford: Blackwell, 2002.

Horgan, T. ‘Compatibilism and the Consequence Argument’ Philosophical Studies 47 (1985) 339-356.

Jubien, M. Contemporary Metaphysics, Oxford: Blackwell, 1997.

Mackie, P. ‘Fatalism, Incompatibilism, and the Power To Do Otherwise’ Nous 37: 4 (2003) 672-689.

Narveson, J. ‘Compatibilism Defended’ Philosophical Studies 32 (1977) 83-87.

Schnieder, B. ‘Compatibilism and the Notion of Rendering Something False’ Philosophical Studies 117 (2004) 409-428.

Taylor, R. Metaphysics (4^th ed.), Englewood Cliffs: Prentice Hall, 1992.

Van Inwagen, P. ‘Reply to Narveson’ Philosophical Studies 32 (1977) 89-98.

Van Inwagen, P. ‘The Incompatibility of Free Will and Determinism’ Philosophical Studies 27 (1975) 185-199.

---

[1] For an example of such a denial, see D. Dennett, ‘I Could Not Have Done Otherwise – So What?’.

[2] P. Van Inwagen, ‘The Incompatibility of Free Will and Determinism’ [henceforth, IFD], p.197.

[3] Ibid, p.191.

[4] B. Schnieder, ‘Compatibilism and the Notion of Rendering Something False’, pp.418-421, convincingly argues that an appropriate interpretation of ‘render’ must involve an explanatory relation. That is, to say that ‘J renders P false’ is to say that ‘P is false because of J’. Sure, we causally interact with the world, not with propositions. But even if the latter are seen as eternal and immutable (see note 11), their truth values can nevertheless be explained by reference to real world states of affairs at some particular time.

[5] As J. Narveson put it in ‘Compatibilism Defended’, p.85, “if J had raised his hand at T, then that would show that either P0 or L is false; but it would not ‘render’ either of them false. It would show that one had been false all along”.

[6] Van Inwagen, IFD, p.190.

[7] Van Inwagen is quite explicit on this point, indeed he later clarifies his definition of ‘S can render p false’ as: “Either p is false or, if p is true, then there is some state of affairs A such at (a) S can (i.e. has it within his power to) bring about A, and (b) A entails the falsity of p” – Van Inwagen, ‘Reply to Narveson’, p.93 (bold added).

[8] Ibid, p.94, “it may be that I have chosen a misleading name for this relation. Well, anyone who does not like my name for it is free to call it ‘Charley’”.

[9] Van Inwagen, IFD, p.192, “if Q entails R, then the denial of R entails the denial of Q. Thus, any condition sufficient for the falsity of R is also sufficient for the falsity of Q.” In the present case, we simply substitute Q = “the conjunction of P0 and L”, and R = “P” into this general analytic principle.

[10] Ibid.

[11] As M. Jubien (Contemporary Metaphysics, p.127) reminds us: “Platonic propositions are entities located outside of spacetime… only entities that are located in time can undergo change”. [N.B. despite initial appearances, this poses no threat to the ‘natural interpretation’ of render – see note 4.] This is the basis of the famous “argument from truth” for fatalism. For a (sympathetic) general discussion, see R. Taylor, Metaphysics, pp.54-67. An explication of the argument is given in Jubien, p.126.

[12] Causation is a ‘one way street’, so to speak. This, I think, is the basis of the past/future distinction.

[13] Especially since he thinks the arguments for it are demonstrably refutable – see Van Inwagen, IFD, p.198 (footnote 6). I suppose he could dodge my objection here if he were willing to give up the ‘law of the excluded middle’, and claim that propositions about the future have no truth value whatsoever (i.e. are neither true nor false).

[14] Jubien, p.128, offers an elegant argument to this effect.

[15] I should note that P. Mackie, ‘Fatalism, Incompatibilism, and the Power To Do Otherwise’ also notices the “striking similarity” (p.672) between the fatalist and incompatibilist arguments discussed here – though she takes her analysis in an entirely different direction than that of mine in this essay.

[16] Throughout this essay, I have understood the ‘hypothetical’ situation to be one whereby J’s different desires were caused by the past being slightly different (i.e. P0 is false, though L remains true). This interpretation leads, as I have shown, to the denial of Van Inwagen’s premise #5. Alternatively, the hypothetical situation may be understood in terms of an identical past which diverges from reality due to a miraculous suspension of the laws of nature (i.e. L is false, but P0 remains true). In this case, analogous reasoning will lead to the denial of premise #6. I have not explicated this parallel argument here, due to space limitations, but it is largely redundant anyhow. A full explication of the analogous arguments provided by the ‘altered-past compatibilists’ and the ‘divergence-miracle compatibilists’ can be found in T. Horgan, ‘Compatibilism and the Consequence Argument’, see especially pp.342-3.