Re: Searle's Chinese Room Argument

From: Harnad, Stevan (harnad@cogsci.soton.ac.uk)
Date: Mon Feb 24 1997 - 21:10:52 GMT


> From: Hawkins, Sean <swh196@soton.ac.uk>
>
> although language may be a form of symbol
> manipulation, this really only applies when it is written, not spoken.
>
> When you read the sentance "No, do not do that" (although out of any
> particular context), you will necessarily believe that the
> object/person was requesting a termination of something, possibly
> responding to a question or instructing another person to disengage.
>
> Now, if the sentance is spoken, emphasis can be given to just about any
> word and depending on which one(s), will or can, for a lay-person,
> change the meaning.

Yes, emphasis can change the meaning of a spoken sentence, but written
language has its tricks too {;>)

> This in some ways can be related back to the
> Chinese Room - the response of the man in the room, may have differed
> if they had heard the spoken "symbol" as well as seeing it.

It doesn't matter whether the input and output is writing or speech. The
conclusion is the same. (Rules can generate -- and modulate -- speech
just as readily as writing.)

> The point I am trying to raise is that although rules of grammar exist,
> they can be broken under the quasi-guise of universe of meaning (ie the
> vast majority understand the meaning of a grammatically correct or
> incorrect sentance based on a whole host of factors eg tonal
> inferences, body language, context).

Stylistic rules are broken ("between you and I," "this is different to
that") but Universal grammar is not. See:
http://cogsci.soton.ac.uk/py104/pinker.langacq.html

> Any true simulation of language must surely therefore have to take
> these things into consideration - any that do not may be simply
> text-book simulations that do not concur with real life.
>
> Are there models that can cope with this idiosyncratic feature of the
> human organism ?

Can a model break rules? Sure: either give it rules to break rules
now and then, or give it some bad rules.



This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:23:50 GMT