Re: Symbol Grounding Problem

From: HARNAD Stevan (
Date: Mon Jun 03 1996 - 16:28:14 BST

> From: "Lander, Emma" <>
> Date: Thu, 23 May 1996 11:20:54 GMT
> Symbols are arbitrary; they have no direct association with what they
> represent so they are not self explanatory.

No direct physical connection. They might be "associated" with
something, but that association has to be noticed and used by a person
with a mind, because in and of itself it is not IN the symbol system,
any more than there are meanings in a book (a book too just consists of
a lot of symbols).

> They only take on the
> meanings that we give them and so they must be 'grounded' into
> perceptual categories which will be easily recognised and understood
> by those using the symbol system.

No, the symbol system itself must be physically grounded in what its
symbols are about; if it depends on the mediation of "users," it is
still ungrounded. The system itself must be able to pick out what its
symbols are about.

> If all you are given is symbols you will not find their meaning by
> looking at more symbols. This is analogous to trying to learn to
> speak chinese from a chinese to chinese dictionary, the symbols are
> meaningless untill you have been told what they represent, and these
> meanings must follow an agreed convention: it is no good having a
> symbol that represents something to one person and something totally
> different to another.

The problem is not that symbols may represent different things to
different people. The problem is that the meaning of the symbols is not
in the symbol system but in the minds of its users. Until the symbols
are directly connected to their meanings without the mediation of
external minds, a symbol system cannot be a model for what is going on
in the mind -- or at least not for ALL that is going on in the mind.

> The meaning of a symbol is in the mind of the interpreter so in order
> to know that a symbol means the same thing to two people an analogous
> image must be provided.

It is not ambiguity that is behind the symbol grounding problem; even in
a grounded symbol system there might be ambiguity; it may nto mean the
same thing you or I do. But at least it would mean SOMETHING.

> For instance, it is no good knowing the
> chinese word for dictionary if you do not know that the word refers
> to a dictionary, you must form a perceptual link between the symbol
> and the object.

You are still thinking of using a dictionary, rather than finding a
mechanical, mindless model for what is going on in the mind.

> The problem is to not ground symbols in more symbols, instead they
> must be placed in perceptual categories.

The system itselg must be able to pick out the members of those
categories that its symbols stand for.

This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:23:43 GMT