Learning, Acquired Dispositions and the Humean Theory of Motivation

Abstract

A central point of contention in the ongoing debate between Humean and anti-Humean accounts of moral motivation concerns the theoretical credentials of the idea of mental states that are cognitive and motivational at the same time. Humeans claim that this idea is incoherent and thereby unintelligible (M. Smith, The Moral Problem, Blackwell 1994). I start by developing a linguistic argument against this claim. The semantics of certain ‘learning to’ and ‘knowing to’ ascriptions points to a dispositional state that has both motivational and cognitive properties: habitual knowledge, as we may call it. But there is nothing unintelligible or incoherent about such ascriptions as they figure in the explanation and assessment of action. This suggests that the idea of a state that has both cognitive and motivational properties is not an artefact of philosophical speculation. Moreover, I suggest that action explanations that appeal to habitual knowledge, which are a variety of habit explanation, present distinctive problems for Humean accounts. The discussion bears on the relationship between habitual knowledge and knowing-how, and its possible significance for anti-Humean accounts of moral motivation.

Acknowledgements

I am grateful to three anonymous readers for helpful comments on a previous version of this paper, and especially for raising the objections and concerns addressed in Section 6.

Notes

1 The classic statement of anti-intellectualism is to be found in Ryle (1949). Stanley & Williamson (2001), as well as Stanley (2011) provide the most comprehensive statement of intellectualism.

2 We have a natural tendency to read sentences charitably, which might lead us to read sentences such as the ones above as ascribing some kind of capacity, a reading on which they are evidently fine. So here it is helpful to use a distinctive mark of the habitual reading: negation scoping over the infinitival, which serves to rule out the capacity reading (Douskos 2017a). This is why [8] serves to make the point more clearly here. Of course, it is easy to provide negated counterparts of [7] and [9]. For instance: [9’] Hannah {has learned/knows} not to drive fast when it rains, {and/but} she does drive fast when it rains. In a nutshell, the reason negation is such a strong mark of the habitual reading is this. Just as one may have cultivated to propensity to a in c, so one might have learned not to a in c, that is, one might have learned to avoid, abstain or refrain from a-ing in circumstances where one used to be liable to do it. This is why habitual knowledge ascriptions have negative counterparts, however pragmatically odd these may be. By contrast, when one is skilled in a-ing, there need not be (and typically there is not) a counterpart skill at non-a-ing: one can be learning, practising or improving at playing the lyre, driving or swimming, but not at non-swimming or non-driving, or non-playing the lyre. Consequently, an ascription ‘S has learned not to a (in c)’ can only be read as ascribing habitual knowledge.

3 Another possibility is that ‘has learned/knows to a’ constructions conventionally implicate a habitual sentence. Failures of reinforceability are usually considered to be evidence against the existence of a conventional implicature (Potts 2007). I am not sure that [10] to [12] are as bad as [7] to [9], so I am willing to allow the possibility that a ‘learning to’ ascription conventionally implicates a habitual sentence. This does not affect the overall argument of this paper. Conventional implicatures give rise to entailments and are generally considered to be a semantic phenomenon at heart (see Potts 2007).

4 For a semantic treatment of disposition ascriptions in terms of habitual propositions, see Fara 2005.

5 These labels are inspired from Thompson’s (2008) discussion habits and practices. As Thompson (2008, 164–165) explains, however, the generality and actuality are features ‘of anything that can be captured in habitual and generic forms of thought’. So the qualification motivational is quite important.

6 I stress this assumption because there is a possible gap in the argument here. The argument will not be valid if the idea of habit is subject to further requirements, and these are not satisfied by habitual knowledge. Nevertheless, I think that the two requirements above capture the core idea of habit, and that any additional requirements will hold with equal plausibility for habitual knowledge.

7 This feature is often called the ‘automaticity’ of habit. Philosophers such as Pollard (2008) and—with some qualifications—Brett (1981), as well as almost all psychologists, take automaticity to be a defining feature of habit, and the absence of deliberation a defining feature of automaticity. I have argued against this definitional claim (Douskos 2017c), but I would not dispute that habitual acts can be, and often are, automatic in this sense.

8 For different accounts of the rationality of habitual acts see Pollard 2008 and Snow 2009.

9 This is why several authors have argued that the causal theory of action cannot accommodate the intentionality, or even the agential character, of habitual acts (Pollard 2006; Douskos 2017b, inter alia).

10 ‘Typically’, because in certain repudiated bad habits, as well as habits that operate below the agent’s awareness, it is not clear that the circumstances are plausibly considered one’s reasons for acting. However, this is clearly so in acts that manifest habitual knowledge.

11 Glick here has in mind the embedded question deontic reading mentioned in Section 2 above. On this reading, learning how to a is learning that, for some way w, w is a/the way one ought to a. So when uttered in a suitable context, the embedded question deontic reading of ‘Hannah has learned how to drive when it rains’ is equivalent to [15]. Therefore the points made below against the equivalence of [15] and [16] entail that the embedded question deontic reading is clearly distinct from the habitual reading.

12 Burnyeat (2012, 265–267) offers a different proposal on this issue. He suggests that ‘S knows that she ought to a (in c)’ is in fact ambiguous. On one reading, it ascribes mere knowledge of a proposition with deontic content, which the agent has not sufficiently appreciated. Hence this knowledge does not serve to improve behaviour. On a second reading, it requires that the knowledge ascribed is experientially grounded on prior habituation. So on this second reading, the ascription is more or less equivalent to a habitual knowledge ascription. I believe that Burnyeat’s concerns are better addressed by the points below, for the ambiguity he relies on has no independent linguistic motivation.

13 The cancellability test is designed to uncover an implicit contradiction. But the two parts of the ascription will be contradictory only as long as the meaning of the implicit ‘ought’ in the first, and overtly expressed ‘ought’ in the second, is the same. The same holds for the reinforceability test. So in checking the non-defeasible inference claim above with examples it is crucial to keep fixed the meaning of the implicit and overtly expressed deontic modal in the two parts of the ascription. The ascriptions [17] to [22] are designed to fix on an instrumental reading of ‘ought’. But in other cases things may be more complicated. Suppose Hannah is very bad at lying. Each time she has tried to tell a lie, she has found herself exposed, so she has developed a propensity to avoid lying. In this scenario, there is still a non-defeasible inference from ‘Hannah has learned not to lie’ to ‘Hannah knows that she ought not to lie’. But the ‘ought’ here is instrumental. Yet Hannah might also know that she ought not to lie for moral reasons. Now consider the following ascription: ‘Hannah has learned not to lie; in fact, she knows that she ought not to lie’. Given the scenario above, the second part of the sentence may be a genuine addition: it says that, in fact, Hannah knows that she ought not to lie anyway, because she knows that this is morally required. Thus the second part is not redundant, but this is only because the meaning of ‘ought’ has shifted. (Notice though that Hannah is not plausibly described as honest in this scenario.) We have a natural tendency to read the examples above charitably, that is, to give different meanings to the two ‘ought’ so as to make sense of the sentence. But we should resist this tendency in order to see the point at issue.

14 As above, failure in the reinforceability test is considered to be evidence against positing a conventional implicature as well. But as I have explained, I do not need to exclude this possibility.

15 Some authors have proposed that ‘knows to’ can only designate a manifestation of knowing-how, as opposed to a standing propensity. The idea is that one knew to a just in case she a-ed in the course of exercising her knowing-how in b-ing, and a-ing was a way of b-ing on the occasion. Thus Hetherington & Lai claim that ‘knowing-to’ is not a generic standing propensity, but only an ‘instance’ of knowing-how (see Hetherington & Lai 2015, esp. 276 fn15). This view has at least three problems: (i) it commits one to the false claim that ‘knows to’ has only eventive readings, since only such readings designate a manifestation of a propensity on some specific occasion; but ‘knows to’ has no such reading in the present tense; (ii) it cannot account for the distinct semantics of habitual ‘learning to a’ ascriptions, which do not allow for a corresponding eventive reading at all; (iii) it cannot accommodate the notion that knowing to a in c can be the outcome of a process of learning to a in c. For this entails that it is a standing propensity acquired by repeatedly a-ing (in c). David Wiggins (2012, 19) allows for a distinction between ‘episodic and dispositional’ readings of knowing-to. But he also seems to think that the episodic (i.e., eventive) reading primary: ‘One who knows at t to V at t is one who makes that direct response to something which the world offers to his perception and which nature, second nature, or habituation and experience invest with a practical significance that demands or suggests or requires such a response (to V, that is)’ (Wiggins 2012, 113). The temporal modifiers ‘at t’ are used to artificially engender an episodic/eventive (occasion-specific) reading, which is not naturally obtainable in the present tense. This is why when Wiggins (2012, 113) gives an example in plain English he has to revert to the past tense: ‘Stubb knew to stay well to the east of the channel.’ ‘Knew to’ here is eventive and implicative. Indeed, Wiggins (2012, 113 fn25) invites us to compare ‘knowing to’ with the implicatives ‘remembering/forgetting to’. So it is hard to see how the formulation above can be amended to accommodate the ‘dispositional reading’. That said, I need not dispute that ‘knew to a’ can also designate a manifestation of knowing-how. But it is clear that this is not always so, for the reasons given above.

16 This is why in using a habitual knowledge ascription a speaker typically expresses or communicates a normative or evaluative attitude. There is something objectionable with ‘John knows to spit at his teacher when he is told to stay quiet’, or ‘John knows to torture the immigrants as they come ashore’. For these sentences convey that the subject ought to be doing the things in question. By contrast, there is nothing objectionable as such in the mere use of the corresponding habit ascriptions.

17 It would not help the Humean to appeal to the intellectualist analysis of knowing-how ascriptions (Stanley & Williamson 2001), for we saw that these are semantically distinct from habitual knowledge ascriptions. Moreover, in Section 4 we saw that ‘S has learned to a (in c)’ is not equivalent to ‘S has learned that she ought to a (in c)’. In the absence of any other plausible contender, we can safely conclude that the object of habitual knowledge is not propositional.

18 As Smith points out, he is engaged in ‘formulating a philosophical conception of folk-psychological states’ (1994, 108; this point is also stressed by Pettit 1987). The dispute is about the conception of mental states that figures in everyday explanation an assessment of action. This makes the examination of the relevant ascriptions all the more pertinent.

19 Since generics and habituals are not universal generalisations, the inference from the generic to the occasion-specific (c) is a defeasible one. It may be true that Hannah (generally) ought to drive slowly when it rains, but on this evening she might be driving a friend to the hospital. Evidently, the inference in the other direction fails as well.

20 On this last point, see Annas 1993, 50–51; 2011, chs 2, 3.

21 For a similar thought regarding how habits work in tandem with skills (‘situation specific discernment’) in virtuous acts, see McDowell 2007, 341.