So why can’t you intend to drink the toxin?

Abstract

In this paper I revisit Gregory Kavka’s Toxin Puzzle and propose a novel solution to it. Like some previous accounts, mine postulates a tight link between intentions and reasons but, unlike them, in my account these are motivating rather than normative reasons, i.e. reasons that explain (rather than justify) the intended action. I argue that sensitivity to the absence of possible motivational explanations for the intended action is constitutive of deliberation-based intentions. Since ordinary rational agents display this sensitivity, when placed in the toxin scenario they will believe that there is no motivational explanation for actually drinking the toxin and this is why they can’t form the intention to drink it in the first place. I thus argue that my Motivating-Explanatory Reason Principle correctly explains the toxin puzzle, thereby revealing itself as a genuine metaphysical constraint on intentions. I also explore at length the implications of my account for the nature of intention and rational agency.

Keywords:

• intention
• toxin puzzle
• metaphysical constraint
• motivating reason
• Mele

Acknowledgments

I would like to express my deep appreciation for the outstanding comments, suggestions, and objections made by the referees for this journal which greatly improved the paper through several revisions.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes on contributor

Fernando Rudy-Hiller is a researcher at the Institute of Philosophical Research (UNAM) in Mexico City. He works on moral responsibility and practical rationality and has published several papers on these topics.

ORCID

Fernando Rudy-Hiller http://orcid.org/0000-0002-7977-1216

Notes

1 A metaphysical constraint on intentions is a constraint that identifies conditions the violation of which makes it impossible for an attitude to count as an intention. In turn, a rational constraint is one which identifies conditions the violation of which makes it impossible for an agent to rationally form an intention, although it may remain possible for the agent to acquire the intention through non-rational processes (by hypnosis, say). In this paper my main interest is on metaphysical constraints on intention, although toward the end I will address some issues about rationality as well.

2 Gauthier (1998a) and McClennen (1990, 227–231) are the most prominent dissenters.

3 Part of what this implies is that the agent isn’t a masochist who enjoys suffering for its own sake, for if we admit this possibility, then the fact that drinking the toxin will make him suffer will be a reason for him to drink. It is ruled out as well the appeal to the independent value that steadfastness has for some people (Kavka 1983, 35). See Sobel (1994, 249) for an argument that this value can make it rational to intend to do something that is, previously to the formation of the intention, irrational.

4 Kavka (1983, 34) also stipulates that the agent in the toxin scenario cannot form the intention by using “gimmicks” such as promising someone that she will drink the toxin, hiring a hitman to kill her if she doesn’t drink it, hiring a hypnotist to implant the intention, etc.

5 Other philosophers who endorse the idea that reasons for intention must be reasons for action include Anscombe (1963, 90), Davidson (1985, 213–214), and Goetz (1998).

6 These brief remarks of course don’t suffice to conclusively refute the RKR account. The important point for present purposes is just that, given the seeming feasibility of intending in response to autonomous benefits, a correct explanation of the toxin puzzle must be capable of accommodating this apparent fact.

7 A related suggestion, due to Michael Bratman (1999), is that considering autonomous benefits in deliberation is acceptable as long as the agent expects that executing the intention will be rational. Bratman calls this constraint on intention formation via deliberation “the Linking Principle” and claims that it – coupled with a “no-regret condition” to the effect that the agent wouldn’t later regret to have stuck with her prior intention – is the key for solving the toxin puzzle (see also Farrell [1989, 288]). Although I think that Bratman’s Linking Principle can be falsified, I lack the space here for making that argument. For present purposes, it suffices to say that Mele’s case of Ted discussed below casts doubt on it, since Ted rationally forms an intention the implementation of which isn’t straightforwardly rational (Mele [1992, 188–189] discusses whether it’s rational for Ted to implement his rationally-formed intention and concludes that it might not be). See also Clarke (2008, 210) for another potential counterexample to the Linking Principle (he calls it “the coherence principle”). In note 34 below I say some more about Bratman’s no-regret condition.

8 See for example Gauthier (1998a, 48) and Goetz (1998, 205). However, Gauthier thinks that, if you form the intention to drink the toxin, then you will actually have a reason to drink it, namely that doing so is part of the best course of action available to you (he assumes that forming the intention to drink and then not drinking isn’t a feasible course of action). For criticism of Gauthier’s test for assessing the rationality of plans, see Bratman (2013).

9 Ted’s case doesn’t violate Kavka’s constraint mentioned in note 4 above, which rules out the use of gimmicks (like promises or assassins) that would give agents normative reasons for actually drinking the toxin. The role of Ted’s peculiar dispositions in the example is to ensure the stability of his intention to drink, not to provide him with normative reasons for drinking. In other words, these dispositions are enablers of his intention without being normative reasons themselves. See note 14 and subsection 4.3 below for more on this.

10 One may object that NRP may be false for Ted but not for ordinary agents like us, and so it may still be the correct explanation of why we cannot form the intention to drink the toxin. The problem is that even ordinary agents may find themselves in a Ted-like situation in which it’s not only possible but even rational to intend to ϕ even while having no normative reason to ϕ. For a case in point, involving the adoption of conditional deterrent intentions, see Mele (1992, 193n.11). More importantly, ordinary phenomena like “desiring the bad” (Stocker 1979) and weak-willed intentions (Wallace 2001) also show that it’s possible to intend to do something for which one lacks (and believes one lacks) normative reasons. I’ll discuss at length the significance of such phenomena for understanding the nature of intention in section 4.2 below.

11 This constraint is negative in the sense that it only requires the absence of a belief about oneself not even attempting to ϕ (see Bratman [1987, 38] for a similar constraint). The metaphysical principle on intentions I present and defend in the next section also imposes a negative constraint on belief, but of a different kind. By contrast, many philosophers have argued that intentions require the presence of certain beliefs, e.g., Hampshire and Hart (1958), Grice (1971), Velleman (2007), and Setiya (2007). In 4.2 I explain why is relevant that a belief constraint on intentions is negative rather than positive. See also note 22 below.

12 Beside Mele (1992, 189), other philosophers have appealed to BATP to solve the toxin puzzle. See for instance Clarke (2008, 211) and Heuer (2014, 313).

13 It’s clear that Mele intends BATP to be read as a metaphysical, and not merely as a rational, constraint on intention (see Mele 1992, 189). Heuer (2014, 313) also construes this principle as a metaphysical constraint, while Clarke (2008, 211) opts for interpreting it merely as a rational constraint.

14 This makes clear why Ted lacks a normative reason for drinking the toxin. Normative reasons justify actions by showing that they are desirable or required from the perspective of a particular normative system, which in Ted’s case would be prudence or self- interest. But when the time of drinking comes, actually drinking the toxin won’t be in Ted’s self-interest at all, since doing so is neither a means for getting something he wants (drinking isn’t required for winning the prize) nor something he wants for its own sake (Ted lacks an intrinsic desire to drink toxin). Therefore, that there is no justification, and thus no normative reason, for him to drink the toxin.

15 By contrast, causal explanations couched purely in terms of, say, neurophysiological processes don’t reveal the agent’s motives for acting as she did.

16 Smith (1994, 104) emphasizes that motivating reasons explain actions in teleological terms, that is, “by making [what] they explain intelligible in terms of the pursuit of a goal.” But Ted’s explanation of his drinking the toxin isn’t teleological in this sense, so the natural question is whether this explanation amounts to a motivating reason. I address this worry at the end of 4.2 below.

17 This doesn’t contravene what Mele (1992, 173) calls “Condition C” on an agent’s winning the prize, i.e. that the agent must be convinced at the time of forming the intention that he has (and will have) no reason to drink the toxin. It is clear that the sense of “reason” at stake here is normative reason, not motivating reason.

18 Why? Because, barring special cases like an agent who is intrinsically motivated to do what makes her ill, ordinary agents will lack (and know will lack) any motivation for drinking the toxin intentionally and so will lack an explanation of the appropriate kind for doing so.

19 Thanks to an anonymous referee for pressing the objection to which this paragraph is responding.

20 Thanks to an anonymous referee for raising these worries.

21 I owe the term “normative vigilance”, as well the phrasing of this sentence, to an anonymous referee.

22 I hope this makes clear why my account of intention isn’t “cognitivist” in the sense of accounts of the sort defended by Setiya (2007), Velleman (2007), and others. Unlike cognitivist views, I don’t think that intentions are a certain type of belief or must necessarily be accompanied by certain beliefs. Rather, as I explained above, in my view the belief constraint on intentions – the constraint MERP imposes – is purely negative: it just requires the absence of certain beliefs. In addition, the conditions on intelligibility and explanation I impose on intention formation are much less demanding that the requirement of self-understanding that, in Velleman’s view, is the “constitutive aim” of action. See Velleman (2000, 2006).

23 Here I put aside intentions regarding the formation of what Elster (1979) calls “precommitments”, by which an agent embarks in a course of action that does involve (to varying degrees) relinquishing control over her future ϕ-ing. The clearest example is that of Ulysses, who orders his sailors to tie him up to the mast so that he can hear the sirens’ song without throwing himself to the water. Notice that this case isn’t a counterexample to MERP, since at t₂ Ulysses doesn’t perform an intentional action at all in refraining from jumping to the water.

24 Once again, I thank an anonymous referee for raising this worry.

25 To avoid confusions, it’s important to keep in mind that my claim that intentions can constitute reasons is restricted to motivating reasons. Many philosophers think that intentions can’t be reasons, but it’s clear that they are concerned with normative reasons (see e.g., Broome 2001). Other philosophers have argued that at least some intentions can in fact constitute normative reasons. See for instance Mele (1992, 181) and Sobel (1994, 249).

26 I thank two anonymous referees for raising these objections.

27 An anonymous referee offered a variant of this case, although with a different purpose in mind.

28 An anonymous referee suggested that the assumption that the same type of action obtains regardless of whether the behavior comes about intentionally or unintentionally is mistaken. That is, one can deny that Ted’s drinking or Laura’s cursing are the same type of actions when they are performed intentionally or unintentionally, despite the indiscernibility of the respective overt behaviors. I take that my argument that Ted and Laura are able to form their respective intentions even in the face of their beliefs about the inevitability of certain overt behaviors works equally well even if the referee is right about this.

29 I take this last sentence almost verbatim from the comments provided by one of the anonymous referees.

30 See the third problem discussed below for further defense of this claim.

31 A famous example of an agent who transforms her alien dispositions (in her case, desires) into exercises of rational (free) agency is Frankfurt’s willing addict (Frankfurt 1988a, 24–25).

32 And they are metaphysically problematic (i.e. problematic concerning the possibility of forming certain intentions) when they crowd out other (rational) explanations for the conduct in question. This isn’t the case in the example I offer immediately in the text, since the claustrophobic person does have an explanation in terms of reasons for always walking to her destinations, e.g., that by doing so she avoids suffering disabling fear and public embarrassment. These reasons, however, needn’t make her actions all-things-considered rational.

33 Anscombe (1963, 94) suggests that it’s possible to intend to do something even while fully believing that one won’t do it. See also Holton (2009, 50). It is a further question, which I will leave open here, whether it’s possible to intend to do something even while fully believing that one won’t even try to do it.

34 A referee worries that my treatment of the puzzle in this paper may obscure the central issue that originally motivated Kavka to develop the toxin scenario, namely to object to Gauthier’s theory of rational commitment according to which an agent can rationally commit to a course of action that provides some benefits to her even when actually executing the plan won’t be in her best interest and may actually be directly against it (for instance a government issuing a threat of retaliation in case of a nuclear attack). On the referee’s view, the adequate way to capture Kavka’s insight in developing the puzzle is through Bratman’s (1999) no-regret condition: an agent can’t rationally commit to a course of action if she fully expects that at “plan’s end” she will regret following through with it (which is plausibly the case both in the toxin scenario and in the nuclear deterrence example). I’ll make two briefs points in response. First, as I said in note 1 above, my approach to the toxin puzzle focuses on discovering metaphysical constraints on intention, whereas Bratman’s treatment focuses on rational constraints. These shouldn’t be taken to be competing approaches; rather, it can be argued that they complement each other. Just as a hint of how this argument might go: MERP isn’t opposed to Bratman’s no-regret condition; on the contrary, it seems that the former partially grounds the latter: if one believes one will lack a motivating reason for ϕ-ing at a certain time, then one can plausibly expect regretting ϕ-ing afterwards (or course, MERP isn’t the whole story about regret, since one can certainly regret performing an action for which one does have motivating reasons, such as in cases of giving in to temptation). Second, it’s true that my treatment of the puzzle focuses on the synchronic aspect of it, i.e. on whether the agent can form the required intention at the time of deliberation, whereas Bratman’s treatment focuses on its diachronic aspect, i.e. on the stability (or lack of it) of the intention to drink in the interval between intention formation and execution. Again, however, these two approaches needn’t be seen as incompatible but rather as complementary.