Monday, November 07, 2005

Accidental Representation

Here's a bunch of interesting questions.

1. Does the involvement of (certain kinds of) luck interfere with representation? Suppose, for instance, that I have a drawing in front of me which, considered as a road map of a small town in Wales, would serve as an accurate guide. But suppose it is purely coincidental that the drawing has this feature - it was produced as a circuit diagram. Does the drawing represent the road layout of this small Welsh town?

2. If accidental representation is not possible, is the kind of accident that prevents representation from occuring analogous to the kind of accident which prevents some true beliefs from being knowledge?

3. If accidental representation is possible, what (if any) are the conditions on representation which prevent it being the case that (e.g.) any two resembling things represent each other?


Andrew said...

I think that the most relevant and interesting sense of 'representation' is that related to 'aboutness'. The sense in which we are interested in mental representation is the sense in which we are interested in how and why thoughts can be about the world. The puzzles relating to how fictional names represent is the puzzle regarding what such names are about.

I don't think it's plausible that the circuit diagram is about the layout of roads in a Welsh town. So I don't think the former represents the latter in an interesting sense. (We can make as if it does.)

I think that the role of intention and convention play a role here that isn't played in many of the true belief/knowledge cases. The most likely cases to follow up involve mental representations, I would hazard. In some influential accounts of the semantics for MRs, the attempt to reduce intentionality to some kind of nomic covariance will rule out lucky representation. So there's a kind of reliability and success built in, in the way that e.g. epistemic virtues build in reliability and success in the epistemic case.

Carrie Jenkins said...

Thanks for these thoughts Andrew - mental representation is exactly what I'm interested in. I think the best account of a priori knowledge has it that we can learn truths about the world by examining our concepts because those concepts are non-accidentally accurate guides to the structure of the world. Sometimes I'm tempted to think they could be accidentally accurate guides (accidentally correlated), and that if this happened we would end up with true but non-knowledgeable beliefs. But recently I've been wondering whether examining a bunch of concepts which are only accidentally correlated with the structure of the world would lead you to believe truths about the world at all, or whether those concepts would just fail to represent the world and hence be unsuitable for framing beliefs about it at all (let alone true beliefs about it).

One thing that's not obvious to me yet is that the kind of accident which would destroy knowledge in these cases is the same as the kind which would destroy representation of the required kind. So it might be that representation requires the absence of certain kinds of accidents, but still, *other* kinds of accidentalness in the accuracy of our concepts can lead to our acquiring true but non-knowledgeable beliefs about the world by examining them.

If I were tempted by some story about mental representation whereby nomic covariance suffices for representation, and some story about knowledge whereby reliability suffices for it, then it would be harder to see any gap of this kind. But I'm not really tempted by either!

Robbie W said...

I think my intuitions would vary depending on the kind of representation at issue. My suspician is that the representational function of maps, writing, spoken language etc are going to depend on factors extrinsic to the representation/situation represented. That might be conventions (e.g. Lewis), or causal conditions on representation (e.g. Dretske), or factors about the intention with which the representation is tokened/interpreted (e.g. Grice). Something, anyway.

I guess a lot of people would think something like that goes for mental representation too. However, it does appeal to me a little to think that mental reps are "intrinsically representational" in some way. One way of making the representation relation an internal one (in the metaphysician's sense) is to make it simply a matter of a relation of resemblance holding between representation and thing represented, or isomorphism between the structure of the representation and structure of the situation represented.

If you adopt such a view on mental representation for those kind of reasons, you won't have the option of turning around and placing causal constriants on meaning, since that would make the rep relation extrinsic again. But it's a very strong position to adopt.

I'm underopinionated on (2), but on (3) I guess you might want to draw a distinction between _an item being a representation_, and the conditions which determine the content of the representation. (Compare Fodor on typing of mental events as beliefs/desires etc, vs. a story about how the content of beliefs desires is fixed). So I'd be tempted to give a functionalist story about what it takes for something to be "used as a representation". As e.g. a red ball just isn't used in the right way, it won't have representational content just in virtue of standing in resemblance relations.

Neil Sinclair said...

I would have thought success semantics, or more generally pragmatism, has some interesting answers to these questions.

Regards, (1), the suggestion is that the important factor in representation is how an item (mental or physical) is used, or translated into action. So if you translate the markings on your piece of paper into certain patterns of action, then the success semanticist would say that the piece of paper represents those conditions under which those actions would be successful. So, roughly, the drawing represents the Welsh town if it translates into how you move about that town. (remembering of course that it can misrepresent). This seems to entail that there is no single answer to what the piece of paper represents - it depends on how it is used. So I guess it could represent a circuit-board aswell.

This will also provide an answer to question (3) - a resembling item will only represent another item if it is used in the appropriate way, that is, has the appropriate connection to actions. (I think that's Robbie's point)

However, though success semantics might work how beliefs and maps represent, I'm not sure how it would expand to things such as artistic representation.

Incidently, Steve Buterfill gave a very good paper on this a couple of years ago - I think it should be out somewhere soon.

(I have no comment about (2).)

Carrie Jenkins said...

Hi Robbie,
You suggest that:

One way of making the representation relation an internal one (in the metaphysician's sense) is to make it simply a matter of a relation of resemblance holding between representation and thing represented, or isomorphism between the structure of the representation and structure of the situation represented.

I take it there'd have to be more to it that that, since a single mental item will resemble many different things in different ways and will most closely resemble itself, but we won't (generally) want to say that it represents many different things or that mental representations typically represent themselves. I wonder how we spell out - and more importantly motivate - these further conditions without bringing in any external-sounding stuff?

Hi Neil,
You say:

This will also provide an answer to question (3) - a resembling item will only represent another item if it is used in the appropriate way, that is, has the appropriate connection to actions.

But you could have (and Simon, for one, does have) a success semantics where what matters is what explains why *actual or possible* actions based (in the relevant sense) on the representation are succesful. I needn't actually use the drawing as a map of Wales for it to represent Wales provided that I *could* use it that way and then my actions would typically be successful because of something to do with Wales.

I don't know how attractive a success semantics would look if it didn't build in some sort of modality along these lines.

Robbie W said...

Hi Carrie,

I was thinking about these things as ingredients for a full account. So one idea, for example, would be that for a mental structure to represent a situation you need both isomorphism between the structures and resemblance between the parts. (Depending on how you spell out "resemblance" between structured entities, this might drop out automatically).

Let me say again: this is hardly a popular view, and likely to have lots of problems. The only reason you'd look favourably on it, I'd have thought, is if you're really set on the representation relation being internal. But it's kind of fun trying to work out how the story would go.

(There must be some discussion of this sort of thing and the problems it faces among early moderns, and early C20 folk. I was thinking of Russell's "Vagueness" when talking about cashing out rep in terms of isomorphism.)

Actually, the point about a thing most resembling itself doesn't worry me so much. If everything else worked, why not say that a structure represents lots of things---both itself (trivially) and external situations. Then we can say that we're only interested in the non-trivial kinds of representation.

If I've got him right, I think Neil and I aren't quite suggesting the same approach to (3). Neil's putting forward a way of assigning content to mental reps (via invoking the content of associated actions). Since arbitrary resembling things aren't standing in the right content-conferring relations, they're not going to represent. That's the same sort of answer as Lewis's functionalist about mental content might give (cf. "Reduction of Mind").

What I was thinking of was a Fodor-style functionalism at the level of mental-state types. For example, for Fodor, a mental item (computationally individuated) has to stand in systematic relations to other mental items (and ultimately, behaviour) in order to be a belief or desire---more generally, in order to be a representation. That story is independent of what you say about how mental items get their content (and Fodor famously goes for a causal story rather than a functionalist one at that level).

So though Neil and I were thinking in the same (broadly functionalist) direction, I think my suggestion had the virtue of neutrality on the nature of representation.