Saturday, 31 October 2009

why reasons aren't causes

Donald Davidson gave us a reason for thinking that reasons had to be causes. Julia Tanney killed off his argument - but reports of its death have been greatly played down. I want to start to put this right, and also to challenge a successor argument by Bill Child.

Davidson starts by acknowledging that the rationalising force of reason-giving - coming to understand something by developing a grasp of how it makes sense - is not to be simply conflated with causal explanation. However what he tells us is that by reflecting on certain examples we can see how the rationalising force of reasons is not enough to explain their explanatory potency. This leads him to the prima facie implausible idea that a reason explanation (I turned on the tap so I could water the geraniums) is, despite its univocary appearance, actually an amalgam of both rationalising and causal powers.

What are these examples? They are cases in which someone might have two (or more) reasons for doing an action, but nevertheless act for only one of them. What is it that makes it the case that the one reason eventuates in an action but the other does not? Causation, Davidson says, is the glue that binds the reason (or the reason's representation in the actor's mind - or the 'primary [set of beliefs and desires] reason' of the agent) to the action.

Julia TanneyDavidson, however, has made a mistake that we need more than rationalisation here. Let us start by acknowledging (as Davidson himself would) that we may act for more than one reason. Given this, if someone does indeed have two reasons for performing an action, the best question to ask (Tanney suggests - if I remember rightly a paper I read about 8 years ago!) is not 'What binds the 'operative' reason to the action?' but 'What renders the other reason inoperative?' When we ask this latter question, what we find is that the latter (inoperative) reason must have been rationally trumped by another goal of the agent. Far from needing an extra ingredient - causality - to hitch operative reasons to actions, what we really need is just a deeper understanding of the justificatory play of reasons.

George wants to go for a walk. Two reasons can be offered, both of which index genuine desires of his. First, he hopes to bump into Georgina who he fancies. Second, he could do with getting a bit fitter. George goes for a walk, and we discern that the only reason for which he acts is his hope that he may have an encounter with Georgina. How do we understand this? By positing causal glue between the fancying and the walking? Not at all. What we need to understand is why George's action is not in fact justified by all his reasons. Then we find out: George also has an essay to write. This trumps getting fit for him - but it doesn't trump meeting Georgina. And the 'trumping' here occurs in the space of reasons rather than the space of causes. It is the rationality of his action given his actual preferences, now revealed to us, that the reason explanation makes manifest.

Bill Child tried another tack for making the case that reasons be considered causes:
For any putatively non-causal explanation of an event, we can always make the Davidsonian point: knowing this story allows us to fit the event into a pattern which potentially makes sense of it; but we are still left wanting to know why the event actually occurred, what made it happen when it did.
Later on he suggests:
it is wrong to think that the Davidsonian argument depends on the idea that there are cases in which an agent acts for just one of two equally strong reasons. The point of the argument is that we need to understand the 'because' in 'She phi-d because she believed that p'. Suppose there is never a case in which S has two equally good reasons for an action she performs for only one of them. It is still true that the mere fact that S's attitudes made it rational to phi does not by itself explain her actually phi-ing. ... we must appeal to causation in order to understand the metaphysics of the relation between reason and action.

...the understanding we get from seeing that reason explanation is a form of causal explanation is a reflective understanding of the metaphysics of this form of explanation; we understand what sort of explanation it is, and
how reasons explain actions.

...perhaps we can add strength to the intuition by showing how naturally we exploit the idea of causality in thinking of the relation between an agent's attitudes and her actions. For example, one way of producing a result is to produce in someone else a motive for bringing it about; by inducing attitudes in you, I can affect your actions and, through them, the world beyond you. It is hard not to think causally of the whole transaction, and equally hard not to think causally of each of its stages. The first stage is clearly causal; when I induce some motive in you, I am evidently affecting you causally. And it is equally natural to think causally of the relation between your attitudes and your actions. And that, perhaps, may go some way to vindicate the causalist's conception of the
explanandum in action explanation.
Well, perhaps. But the old-fashioned Wittgensteinian anti-causalist is hardly going to be moved by an approach which simply assimilates actions to events (rather than holds out for an analysis both of them, and of their relevant accountings-for, as sui generis in character), which supposes from the start that explanations of actions are interested in accounting for just why they occurred when they did (perhaps reason explanations are not that specific!), which also denies from the start what the Wittgensteinian suggests: that putting an action (which might not best be thought of as an 'event' or happening) in the context of a sense-making pattern precisely provides us with knowledge of why it was undertaken (if not of why it 'occurred'; similarly we wouldn't expect to explain why events are undertaken since it's only actions which are undertaken), and which in the process construes the art of rational persuasion as a matter of the 'production' or 'induction' of attitudes in someone else. (Although the point of the Wittgensteinian's analysis, it ought to be said, was not to issue a fatwa on the use of causal rhetoric in action-intentional context (although for the causalist to use it to vindicate an argument is a bit rich!). It was rather to suggest that a.) assimilating reasons and causes obscures more than it illuminates, and b.) that some of the reasons why we may feel compelled to bother offering causal analyses in the first place have (historically at least had) more to do with the unwarranted assumption of an estranged conception of the relation of (a disembodied) mind and (a deanimated) body in the first place, a relation which then will seem to be in need of causal glue to link together what the analyst has unwittingly sundered.) Further, no-one (so far as I know) has suggested that the mere fact that it would be rational to act for a certain reason makes it the case that someone acts for the reason in question (the reason needs to be theirs).

What then makes it the case that what could be a reason for someone to act is in fact one of the reasons - one of their reasons - for their action? In answer to this question the anti-causalist is typically happy to cite the kinds of factors which the causalist themselves would, if they were not beguiled by estranged and decontextualised conceptions of mindedness and embodiment, presumably also want to cite - such as the agent's ownership of the reason (do they cite it when asked? for example), or the meaningful contextual character of the situation of their action (the arm raising occurred in the context of a bicycle turning manouvre - John raised his arm in order to signal his exit from the main road). All they hold on to is the (alleged) fact that adding causation to the analytic mix is de trop.

This, the anti-causalist will claim, is just what it is to 'understand the 'because' in 'she phi-d because she believed that p''. We receive the kind of illumination we get when we see what the action in question really is: what the smaller fragment which eluded our prior comprehension is a fragment of. (If someone asks: "Well, and why is that illuminating?", then I think I'd just have to say the same about causal explanations: "Why are they illuminating?" The truth is: these just are what amount to two modes of understanding here: we need to understand what it is to 'understand something' in terms of such particular explanatory endeavours, rather than take it that we possess some prior notion of what 'understanding' consists in which can then be wheeled out in questions such as "And why [what do you mean "why"?] does this count as 'understanding'? ) Causal explanations and reason-giving explanations simply are - on the Wittgensteinian position I am recommending - to be considered two separate sui generis forms of explanation, and there's little sense in analysing either in terms of the other. If however what it is to understand a 'because' in either context is to apprehend something more than the differences between these conceptual contexts - for example, to be able to provide a reductive analysis of the concepts bound up in one form of explanation in terms of quite different concepts - then the anti-causalist of a Wittgensteinian bent will simply reject the project ab initio. Why should we even think such an analysis was necessary or desirable?

Wednesday, 28 October 2009

campbell's causal control criteria

John Campbell gave a talk last night at the Oxford Philosophy Faculty. He talked about the nature of causal relationships, with particular focus on psychological causation. Against those who view causation purely 'extensionally' as one might say - as a fact about the relationships between (say) microphysical aspects of the world, relationships which may be very complex and leave quite unsatisfied our demands for straightforward and simple causal understanding - Campbell seemed to tie the nature of causality itself to that which is explained by satisfying causal explanations.

In particular Campbell explored the nature of 'control variables'. The idea here is that we can rightly claim to have identified the cause of something (e.g. what causes the volume of a radio to increase is the volume knob being turned) when the explanans stands in the relationship of a 'control variable' to the explanandum. I don't recall all of the characteristics of these 'variables', but they include the notion that there should be no gratutitous redundancy, a total mapping of causes onto effects, a dose response, and a ready computability of the cause, etc., in bona fide causal explanations.

Upshots of Campbell's view include the idea that there can, even in a 'deterministic physical universe' be non-physical (i.e. psychological) causation (since the best explanation of most human behaviour is psychological rather than physical: psychological factors behave more like control variables for behaviour than do microphysical brain states). This, to say the least, challenges in a very interesting way some of the presuppositions of mainstream physicalist philosophy.

An obvious worry - that was reflected in at least two of the questions (I had to leave after a few of these) - with Campbell's position is its apparent 'anthropocentrism'. One way to put this worry is to say: Campbell seems to be trying to derive a metaphysical conclusion about the nature of causation itself from explanatory considerations - about what we (humans) find explanatorily satisfying. This is a move which many hard-nosed metaphysically-minded sorts will want to resist.

Despite the worry, I find myself attracted to Campbell's position - mainly because of the easygoing, unpretentious way it pricks the side of the 'oh-we're-so-hard-nosed-and-scientific' contemporary physicalist philosopher of mind. And so I naturally wonder what a good defense of what may appear to be its anthropocentric bias might be.

I found myself thinking of the ways in which Wittgensteinian conceptions of the relation of meaning to use, or of the nature of the will, or of the relation of perception to its objects, have been defended against the objections to them lodged by Grice et al. The Wittgensteinian attempts to draw conclusions about what is intelligible - conclusions about meaning, what it makes sense to say - from premises about what people would say in various situations. For example, they would say that there are many situations in which we would not describe someone as trying to perform some action (because it's so easy and they just succeed straight off 'without having to try'). This then is used to count against theories of action which suggest that intentional action just is, at a minimum, action which we are trying to do.

The Gricean says: 'Well, just because we wouldn't it doesn't mean it's not true. Perhaps we don't say it because it's just too obvious - it goes without saying - it's already implied in the conversation.' Against this the Wittgensteinian (e.g. Glock in his Wittgenstein Dictionary, p. 389) may claim that it's hardly less obvious to say that John is trying to tie his laces when he struggles and fails than when he effortlessly succeeds. The fact of his trying is in truth far more patent in the former case.

Just as the Gricean seems to want to extend talk of (say) 'trying' to a whole gamut of cases - regardless of whether we would normally find it helpful to locate trying there - to the end of providing us with an objective account of what makes for intentional action, so too the metaphysician seems to want to extend talk of 'causation' to a whole gamut of cases regardless of whether we would normally find it helpful to cite the cited events in explanations, all to the end of providing an account of what makes for the happening of happenings. In both cases they find themselves tempted to say "In restricting your talk of 'trying' or 'cause' to those cases in which talk of causes or of trying is genuinely informative, you are merely demonstrating your anthropocentric bias."

By contrast the Wittgensteinian could reply: "The burden of proof lies not with I but with you. You are assuming that I am being anthropocentric (letting my sense of what counts as a good explanation get in the way of my ontology), but I am claiming that the concept just is like that. You are trying to remove it from the context which gives it its life in the first place. It's not that I'm giving a biased account of what a cause can be. It's that you are giving a misleadingly broad account of causation. A cause isn't to be defined as that which features in a satisfying causal explanation (that would just be circular), but we can look at which explanations are and are not successful in specifying causes to give us a good sense of what a cause is. (After all, as I seem to remember Campbell saying, causes just are that which are cited in causal explanations.) In your desire to secure the objectivity of genuine causality you try to distance it from human concerns, but unwittingly thereby also uproot it from its sense-conferring context of application."

Campbell's homely account of causation, to the extent that I understood it, brought that concept nicely back down to earth for me. That process was also helped by Wittgensteinians such as Anscombe - who talk of the ways in which pushing and pulling and cutting and startling are causal notions which jointly determine what is meant by causality (rather than instantiating some further hidden connecting phenomenon called 'causation'). As a result I find myself less inclined to turn to some special discipline called 'metaphysics' to help me work out what that mysterious causal relationship 'really' consists in. Causation is now even more than ever just one of hundreds or thousands of (albeit abstracted) concepts jostling around our richly elaborated form of life.