Wednesday, 11 February 2015

Tolerating approximation in philosophy

At the end of a post-work beer with my colleague Peter Lucas, during which we had mainly discussed the connection between Hume’s problem of induction and quantitative social science research methods, we found ourselves in agreement on a pedagogic point. A too bleak presentation of the force of Hume’s argument would likely leave social scientist students or colleagues feeling somehow cheated, short changed and uninterested. A too rosy presentation of the prospects of dissolution would undermine the lessons Hume offers for the difficulty of mapping real world cases on, say, hypothetical finite jars of differently coloured marbles.

On leaving, unfolding my bicycle to head to the station, I asked him, for interest, whether he had a favourite principled or philosophical solution to the problem and he remarked, lightly, that the principled problem did not particularly bother him. Well why should it? After all, it cannot really engage one (as Hume, and in a parallel case Descartes, both acknowledge) once one has left the study or lecture hall.

Still, on the train home, I realise that I do have a shopping list of solutions to key philosophical problems. I would not be happy to be without some favourite approach to Hume, for example. But this in turn makes me realise that I tolerate a remarkable number of counter examples or unresolved difficulties in order to think that I have at least a dim view of an approximate solution. This tolerance resembles the Kuhnian idea that a dominant scientific theory is born ‘refuted’ and soldiers on despite this. But it doesn’t seem very philosophical. (Such a relaxed attitude would, eg., undermine the nature of question sessions at philosophy presentations.)

To take one example. Donald Davidson argues against type type identity theories in the philosophy of mind by arguing that mental state types, bound by the constitutive ideal of rationality, cannot be identical to physical state types, which are not so bound. John McDowell offers a similar argument against functionalism. Mapping mental states, again bound by the constitutive ideal of rationality, onto functional states would require the codification in functionalist terms of the demands of reason which seems implausible. (McDowell talks here of proof that it cannot be done.) In his book on Davidson, Bill Child deploys a related argument against even a network of token mental states identified with token physical states to which Davidson subscribes. Even that identity requires a parallel codification of the token states against which Davidson’s master argument still applies. Further, it now seems a mere coincidence that the structure of relations governing the mental tokens keeps in step with the structure of relations governing the (very same) physical tokens.

In retreat from these stronger positions, I have always assumed that weak supervenience (in the sense of John Haugeland’s article of that name as requiring no identity of elements rather than the modal version (Kim) as meaning applying only in a narrow range of possible worlds) is a place of refuge. But there is still a related objection: without a principled codification of the structure of the space of reasons in the concepts of the realm of law, or vice versa, why should these two domains march in step? A token identity theory seems unnecessary to raise this worry. I tend to ignore this, relieved that I do not publish in this area, but still retain some reassurance that despite the objections the solution lies in this broad area (supervenience without identity).

But on sober reflection, that seems rash. The objection looks to be a haymaker: ruling out occupying this possible space in realm of possible solutions to the mind-body problem. Of course, as a cursory glance at this ever-getting-another-job-defying blog would reveal, the same applies in other areas. I often flag philosophical commitments whilst simultaneously advertising obvious objections to them.

I would like to offer as a response to the question I set Peter a broadly externalist but reason-based disjunctivism about induction (sketched here and here). But I cannot swear that I lack misgivings about addressing the objections it prompts.