From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!bonnie.concordia.ca!garrot!uxa.ecn.bgu.edu!ils.nwu.edu!pautler Tue Nov 26 12:30:54 EST 1991
Article 1436 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!bonnie.concordia.ca!garrot!uxa.ecn.bgu.edu!ils.nwu.edu!pautler
>From: pautler@ils.nwu.edu (David Pautler)
Newsgroups: comp.ai.philosophy
Subject: semiotics and cognitive science
Message-ID: <4219@anaxagoras.ils.nwu.edu>
Date: 20 Nov 91 02:52:32 GMT
References: <1991Nov11.024611.12312@nuscc.nus.sg> <LmVaBB1w164w@depsych.Gwinnett.COM> <1991Nov14.065924.29076@nuscc.nus.sg>
Sender: news@ils.nwu.edu
Reply-To: pautler@ils.nwu.edu
Organization: The Institute for the Learning Sciences
Lines: 53
Originator: pautler@aristotle.ils.nwu.edu

In article <1991Nov14.065924.29076@nuscc.nus.sg>, john@mango.iss.nus.sg (John Waterworth) writes:
> 
> The code-decode model is simply the idea that, you know, I have this
> notion in my mind which I code into some linguistic representation which
> I convey to you, in speech or writing, and which you then decode to get
> at the notion I had in the first place. There was a lot of work on the
> kinds of mutual knowledge needed for this to happen. But then it seemed
> that the adjustments to the model became so complex that it was better
> abandoned for a more parsimonious account.

> Inference can be applied to all kinds of cues (like the height of my
> eyebrows). The inference-ostention approach emphasises that our
> inferencing is always guided by considerations of relevance and context.
> This is partially motivated by a desire to account for irony, sarcasm,
> etc.. where the coded form deliberately belies the communicated meaning.
> The best reference to this, with lots of good examples, is 'Relevance'
> by Dan Sperber and Dierdre Wilson (Oxford: Blackwell: 1986?). It's a key
> work for anyone interested in communication, NLP, AI etc.. Read it and
> you will lose your interest in semiotics.

You know, I find it difficult to disagree with the conclusion here given
its basis in actual attempts to model the communicative process.  But I do
believe that semiotics has something essentially correct about it.  That
something is its concern with convention (or mutual belief).

I'm not sure what work John was referring to, but I see a need in much
of current cognitive science for models that can infer mutual belief from
limited evidence.  The work in analogical reasoning, for example, needs a
better account of how the boundary of an analogy is determined (e.g. if I say
"Bob is a pig", should you think I meant he has disgusting personal habits,
or that he has a short curly tail?).  Theorists in that subfield readily
acknowledge the frozen, conventional nature of metaphor, but don't seem at
all interested in offering an account of how "frozenness" develops or is
recognized.

The inferential power needed to do this *will* be great, but without it,
what hope have we of doing speech acts any justice?  I have not read the
Sperber and Wilson book, but I'm highly suspicious of any theory that claims
to account for relevance or context without relying in some way on mutual
belief.

Semiotics is flawed, and Eco's book was disappointing to me, too.  But one of
the things cognitive science needs right now is an inferential account of
mutual belief (Yorick Wilks' recent "Artificial Believers" may be the right
first step).

BTW, I bet more than half of everything people have ever written has its
basis in conventions -- all the discussions of meaning, arguments over
moral justification or compliance with the law, retelling of myth and
legend, wrangling over philosophical truths, etc.  The difficulty with
convention or "culture" is that it is so easy to confuse with "nature".

	-dp-


