Tuesday, 31 August 2021

Pickering on family resemblance

There’s a discussion of family resemblance in Neil Pickering’s 2011/12 paper ‘Extending disorder: essentialism, family resemblance and secondary sense’. Given that family resemblance trades on resemblances which are simply likenesses and given Pickering’s critique of the ‘likeness argument’, I’m surprised that he is as positive about family resemblance as an account of illness. It encourages me to rethink the conclusion of his Metaphor of Mental Illness as reinstating likenesses but only via their metaphoric construction. In other words, conditions might indeed count as illnesses via likenesses with paradigm illnesses (bizarrely in accord with the likeness argument) but only given prior creative mis categorization of their characteristics.

Anyway, this paper has what looks to be a critique of family resemblance concepts (FRCs) which is later ‘turned on its head’ and used to highlight a virtue of seeing illness as a family resemblance concept.

One confusing feature of the paper is that it uses the word ‘extension’ when discussing concepts but not generally to mean the class of things to which a concept applies (the logical sense of the word). Instead it is more often, at least, the noun formed from the verb ‘extend’. But I wonder whether this potential ambiguity causes Pickering himself some momentary confusion. That it means the latter is suggested by the framework of the paper.

Extending a concept refers to increasing the range of things which can be brought under the concept, or which fall within a category, or to which a kind term can be applied. The increase is from a baseline which is the agreed range of usage, reflecting what Wakefield calls ‘‘relatively uncontroversial and widely shared judgements’’ about what does (and does not) fall under the concept.

Two conditions are then outlined for such extending: determination and coherence.

An extension should neither go too far beyond the baseline and include things which should not be included within the concept; nor fail to go far enough beyond it, and exclude things which should be included under the concept… Echoing such concerns Horwitz suggests that to be an ‘‘adequate concept of mental illness’’ a concept must demonstrate the capacity to ‘‘distinguish conditions that ought to be called ‘mental illnesses’ from those that ought not’’… The criterion of valid extension implied by Bellaimey, Horwitz, and Wakefield will be named the criterion of determination.

[I]ntuitively, random lists of items do not imply the existence of any covering concept. What, for example, is the concept implied by (square root of) 9, the chair I’m sitting on, Beijing, and the Mona Lisa? What is lacking from this set of items is any sense that they belong together. There is a need for a criterion of coherence.

This argument for the second condition goes too quickly. There seems no reason to preclude this arbitrary list as instancing a general concept ie having Evansian generality. Let’s call it: Pickering’s example of random things. We can think of the list under that generality. Under counter-factual conditions, different objects might have occurred to him.

Pickering describes FRCs thus:

The family resemblance approach can be represented schematically and abstractly... Five things (labelled here A, B, C, D, E) are agreed to fall under a concept—they form the baseline of agreed examples. There are also five characteristics or attributes the things collectively have (labelled I, II, III, IV, V).7 Each of the things has four of the characteristics, but each has a slightly different set:
A [I, II, III, IV]
B [I, II, III, V]
C [I, II, IV, V]
D [I, III, IV, V] E [II, III, IV, V]
Consistent with the main point Wittgenstein seems to be making, there is nothing these five things all share.

Taking these to be the baseline concept, Pickering then considers a new case:

Now, let it be asked: does a further thing (F) which has the following characteristics, also fall under the concept in question:
F [III, IV, V, VI]
This shares at least three characteristics with D and E. If, on these grounds, this is allowed to be a new instance of the concept, a further characteristic (VI) may now be considered a characteristic of things of this kind. And so the characteristics which are associated with the concept are increased in number and range.

Before mentioning the disastrous logic of this example, let’s continue with Pickering’s application of the schematic model to a mental illness case.

Despite its name, the medical status of ASPD is contested... As described here, this condition has no overlap with the existing set of characteristics associated with the notion of disorder, and so on a family resemblance account there is no reason to bring this instance under the concept. However, this can change. There is reason to think that there is a subgroup (G*) amongst those exhibiting antisocial personality disorder who also respond to medical treatments (III), are typically harmed (e.g. suffer a sense of social isolation) (IV) and have a genetic predisposition (V)… If sharing three attributes or features with things which are agreed disorders, is sufficient to make something a disorder, then G* will be a disorder, and since the gap between F (schizophrenia) and G (anti social personality disorder) has now been bridged, it too will be a disorder.
In the light of this, a family resemblance account of disorder does not appear to meet either criterion for extension. The attributes don’t appear to determine any limits to extension, since they seem to be an indeterminately extendable list. Further, A and G are now included, yet they share no attributes, and so it is unclear what the rationale for including both might be.

In fact, Pickering goes on to defend the FRC view of illness. But it is worth emphasising just how radical this objection would be if it were plausible. Pickering’s initial schematic example concerns F. But there is no reason to assume that F is actual. We can explore conceptual generalities through merely possible instances. So if this case extends the concept, we can imagine a range of further extensions via a range of further hypothetical cases. Assuming, following Pickering’s words here – whether or not he means them himself – ‘a further characteristic (VI) may now be considered a characteristic of things of this kind’ and then nothing seems to be excluded from the extension (in both the logical sense and the sense of extending) of the concept.

There is, however, an assumption built into Pickering’s consideration of this objection in the comment ‘a further characteristic (VI) may now be considered a characteristic of things of this kind’. Everything depends on the ‘may’. It is that that permits the extending of the original concept to anything we like. But it seems to be motivated by this earlier comment when introducing F: ‘which has the following characteristics’. Pickering assumes – on behalf of this objection which he will go on to reject but without actually diagnosing the problem – that what is characteristic of F is characteristic of the kind or concept, too. But that does not follow. To take the other famous example of FRC – other than an actual family resemblance such as the Thornton-look – a specific game might have a characteristic without it being a characteristic of games. Golf has as a characteristic the fact that it is played by the wealthy. That’s quite typical of golf. But it does not infect games as such. So the putative objection depends on an implausible assumption.

Pickering himself rejects this objection saying that ‘the argument so far considered against the family resemblance approach must be turned on its head’. The idea seems to be that FRCs balance strict normative determinacy with an element of flexibility. One motivation for this is to argue that even non-FRC and essentialist accounts permit some flexibility. He says:

Even if the essentialist were correct that there is a pattern of necessity and sufficiency in the features of those things which are agreed to be disorders, it would be a further decision to conform the notion of ‘disorder’ to that. This decision can’t be justified by the pattern of likenesses itself—since it is the implications of this which are in question; rather it must reflect a human decision or practice.

But is that right? Consider Wakefield’s claim that a disorder is a harmful dysfunction. Suppose we accept that definition, we have necessary and sufficient conditions for calling anything a ‘disorder’. There is no further slack aside, say, from what word shape or sound we attach to the concept. Things are disordered in virtue of being harmful and dysfunctional. The conditions do not merely group conditions – extensionally, as it were – but group them as disorders.

Of course, if we just have the extension of disorder, there would be a further question of in virtue of what these things are conceptually united. But Pickering has already notionally conceded more: the necessary and sufficient conditions for being a disorder.

(I’m not sure about this, but I wonder whether Pickering’s account FRCs shares something of this picture. The putative objection to FRCs seems to start from a position outside the conceptual order, trying to infer the concept that unites a collection of otherwise unconnected instances. That is why, I hypothesise, FRCs seem so hostage to the fortune of subsuming a new instance under them. There is no sense of in virtue of what items are grouped together. (Hence in my example, being played by wealthy people might get attached, willy nilly, to ‘games’.) It is an alienated picture of our access to our own thoughts. As though we have to run a kind of Davidsonian Radical Interpretation of our own linguistic dispositions to work out the conception that gathers together the extension (in the logical sense). Again, this makes more sense if one recalls the idea that in his paper ‘extension’ generally means an extending but may also mean the logical sense of the class of things to which a concept applies. This may become clearer below.)

Anyway, Pickering’s response to the objection above runs thus:

But these problems for the family resemblance approach are more apparent than real. They have arisen as a result of focusing on the role of the attributes or characteristics of things in deciding category or kind membership. This focus on the characteristics represents a mistaken approach to family resemblance. Characteristics in the family resemblance approach to concepts do not have the determinative relation to the concepts with which they are associated that they have in the essentialist approach. Furthermore, this is a strength of the family resemblance approach, for the kind of relationship the essentialist presupposes between characteristics and concepts is far from self evident…

The possibility inherent in family resemblance of continuously extending the features associated with a concept indicates the existence of a non-determinative relationship between those features and the concept they are associated with. It is just because networks of resemblances exist between quite different kinds of things that the family resemblance approach necessarily introduces a further factor. In the family resemblance approach it becomes clear that some judgement has to be made about whether shared attributes do or do not count towards various human patterns of behaviour and experience being disorders…

Answering these questions provides a basis for a boundary of the extension of the concept, and so meets the requirement for determination. Where it is judged that resemblances of attributes count for nothing, then the limit of intelligibility in the application of the concept has been reached.

Here I think we have the crux. Indeed it applies to interpretations of Wittgenstein’s discussion of rule following outside FRCs too (eg Kripke, Travis, Wright). Does Wittgenstein’s criticism of platonism, of the idea that the conceptual order is utterly independent of us, leave room for, or indeed require, that we make things up? The slippery notion in the passages above is ‘judgement’. If I judge that 7 is greater than 5 by an increment of 2, I judge it but I judge it correctly. Any other judgement would be wrong. I judge in accord with the rules of maths. But there is another sense of ‘judgement’ in which the idea of correctness or accord with a standard goes missing and the judgement itself sets whatever lesser sense of a standard there is. The score - on Pop Idol, say - is created by the judgement of the judges, as it were.

Pickering wants family resemblance to have a foot in both camps. In the passage concerning the possibility inherent in family resemblance above, it sounds as though we get to make up the extension of the concept (in both senses of extension) but in the next passage - which is the very next paragraph but with a couple of sentences omitted by me - it sounds as though we answer questions which have antecedent correct answers and so there is genuine normative determination.

I’m not sure he can have it both ways. My suspicion is that this is where the ambiguity of the word ‘extension’ applies. Extending a concept changes the concept. Subsuming a new instance under a concept need not (pace Travis). Normative determinacy applies to the latter, only.

I wonder, too, whether this has something to do with Pickering’s idea that there are limits to the application of even FRCs. Like counting to 5 for Monty Python’s Holy Hand Grenade, some applications are ‘right out’ or, in this case, simply unintelligible. This, however, seems like a final Carrie-like grip of Platonism from beyond the grave. Conceptual normativity is explained by an appeal to what is unintelligible. But surely, the unintelligibility of the application of a concept to a range of cases flows from the concept involved? No deployment of symbols can be brutely unintelligible without reference to the rules given to the use of words and hence a concept. It cannot explain the limits of that concept. That would be Platonism.