Future Research

September 2, 2011

This is not a full return to blogging – just putting some thoughts up here to clear my mind. Picking up where I left off.

Brandom’s apparatus allows us to understand how conceptual and normative content can be produced by nothing more than a complex interactive system of reliable differential responsive dispositions. That is to say: Brandom’s apparatus makes a major contribution to the naturalistic project of explaining how norms and concepts can emerge from natural- and social-scientifically analysable behaviour.

Of course, there is a huge amount that is ‘black-boxed’ in Brandom’s work. And it is likely that Brandom’s wrong about lots of stuff. This is science, after all: we only ever have ‘best hypothesis’. Brandom’s work is a contribution to the ongoing natural- and social-scientific naturalistic project. There is, always, more more more work to be done.

One direction in which more work could be done: the natural-scientific direction of examining how the organisms that participate in the systems of social practices analysed by Brandom actually function. Lots of people are doing this work right now, obviously.

Another direction in which more work could be done: the elaboration and testing of the claims about linguistic practice which Brandom makes. The community of analytic philosophy seems to be picking up on this, in part.

A third direction in which more work could be done: the attempt to apply Brandom’s ideas to artificial intelligence, by using a Brandomian framework to imagine or create algorithmic systems of reliable differential responsive dispositions capable of engaging in the kind of behaviours Brandom discusses. (This is not my bag at all, but again I think people are working on it.)

A fourth direction in which more work could be done: we could attempt to construct an alternative apparatus, parallel to Brandom’s but distinct from it in many details, that could account for the emergence of normative and conceptual content out of non-linguistic practice. My own view, as I’ve discussed before at some length, is that such a thing could be achieved, and I’d be really interested to see it. Of course, I could be wrong about this, and it isn’t a research direction I myself plan to pursue.

A fifth direction in which more work could be done: the addition of non-linguistic practice to a broadly Brandomian analysis of the determinants of normative and conceptual content. Or, put another way, the integration of a Brandomian analysis of the emergence of normative and conceptual content with other accounts of the emergence of normative and conceptual content that do not emphasise the linguistic.

All these are of course related to ongoing research projects that will chunter away quite happily without any input from Brandom or Brandomians – I’m just pointing out some of the connections that can be drawn between Brandom’s work and other research.

The last direction – the fifth – is the one I want to begin to contribute to.

I’ll discuss that in my next post.


2 Responses to “Future Research”

  1. Jed Harris Says:

    Great to see you back on this, even if only occasionally.

    I think your third and fourth research agendas are closely related, perhaps becoming identical as they are carried out.

    “Good old fashioned” AI, pursued through the mid-80s, tried to get computers to replicate human practices by making those practices explicit ways more or less consonant with current analytic philosophy — capturing the practices using appropriate descriptive schemas and interpreting the descriptions. This failed. Explicit descriptions turned out to be fragile, and if they were elaborated to make them more robust, they became unmanageably complex. It appeared that even simple practices were unboundedly rich.

    Note that this problem played out pretty much the same way in both linguistic practices (paraphrasing, translating, extracting actionable requests and answers, etc.) and non-linguistic practices (stacking blocks, navigating across the room without bumping into things, recognizing faces, etc.).

    This problem was pretty much solved by moving to sub-symbolic computing (for example “neural networks”, but these are just a metonym for a large class of statistical techniques). Now we have adequate (though not human level) machine implementations of speech recognition, translation between languages, driving cars, piloting helicopters, etc. To some extent these practices are programmed but they must largely be learned from examples, effectively imitating skilled human practices.

    So two experimental results are (1) human practices can’t be directly encoded in any reasonable humanly comprehensible language, they depend on large amounts of sub-symbolic information and processes, and (2) human linguistic practices are practices in this sense — they depend heavily on sub-symbolic information and processes.

    Which brings us to Brandom. Human practices of framing, using and modifying concepts certainly fall under this general conclusion about human practices. For example categorization of news articles is a basic experimental paradigm in “statistical learning”. Recent work involves forming “good concepts” from a diverse range of examples with little or no guidance.

    AI didn’t try to replicate practices of normative judgement, but the framework you’ve described seems generally feasible given the direction AI is going. As with these other practices, actual practices of normative judgement will certainly largely impossible to capture in language, but they can be entirely “naturalized”.

    I haven’t read your next post yet but I’ll conjecture that the specific ideas AI developed to succeed are likely to be relevant.

    A side note: I wasn’t familiar with the word “chunter” but looking it up, it seems like a somewhat uncharitable term for how these other research programs are operating.

  2. duncan Says:

    Hi Jed – thanks :-). On the “chunter” issue – well, there you go… I did as you did, and googled it, and discovered that my usage is evidently non-standard. I see that Merriam Webster has “to talk in a low inarticulate way”, which seems to be the consensus. But I’ve always used the word to mean something like “slow but steady movement”. To me it’s in the same associative space as “trundle” – if one is chuntering, one is not racing, but neither is one crawling – it’s somewhere in between. In my head it’s onomatopoeic for the noise I imagine an old steam train would have made in motion. [And is generally used fondly.] This association may not be completely idiosyncratic, since when I google around, a lot of the usages that seem similar to mine relate to trains. E.g.:

    Well, I was thinking about all this as my three wonderful friends and I were chuntering along on British Rail across the length and breadth of England last Friday


    Or does Paul travel more ‘Lenin-style’ in something steel plated, fortified with blacked out windows, chuntering along slowly beneath the Atlantic, plotting and co-ordinating his own curious revolution?

    So I don’t think this usage is wholly idiosyncratic, but it’s presumably something I ought to cut down on, since the word evidently means something else to many people…

    In any case, I wasn’t meaning to be derogatory by describing research programs as “chuntering” – the intended meaning of the sentence was just that these programs are doing just fine, and will continue to do just fine, without Brandomian input.

    Will reply to your other points soon…

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: