Sent Mon Feb 12 22:11:03 GMT 1996
Newsgroups: sci.psychology.consciousness
References: <9602090141.AA07993@adage.Berkeley.EDU>
Subject: Re: original intentionality

Stanley Klein <klein@adage.Berkeley.EDU> writes:

> Date: Thu, 8 Feb 1996 17:41:08 PST
>
> Pat Hayes in his critique of Searle, seems to be saying that computer programs
> have original intentionality. Did I get that wrong? If not, I'd like to
> know which program (or operating system) has intentionality without the
> need for an outside observer.

I think the assumption that intentionality is some sort of
all or nothing package fails to do justice to the complexity
of human intentionality. Stan seems to be making that
assumption,  but I may have misread him. If I have, the rest
of this message is irrelevant to his question.

I can't speak for Pat, but I once tried to clarify this sort
of question in an article (in the philosophy section of the
International Joint Conference on AI, UCLA 1985) called
"What enables a machine to understand?"[*]

I tried to separate out some of the strands in the notion of
intentionality (or understanding a notation or
representation) showing that some of them already occur in
dumb computers without any AI software.

Modern computers are inherently semantic engines: otherwise
they would be useless. E.g. they use bit vectors to REFER to
locations in their address space, they use bit vectors as
COUNTERS when enumerating items in their memory, etc.
Slightly more sophisticated systems use tables of records of
recent events as A BASIS FOR DECIDING which processes should
run next. A bit-vector in such a table would be interpreted,
by the operating system, as REFERRING to a process-record.

The process-record would have other structures in it,
interpreted by the machine when it re-starts the process,
e.g. items referring to the procedures currently active in
the process, other items referring to the particular
instructions in those procedures that have to be run when
control returns to the procedures.

It's true that machines do this because we made them do it,
but the more important point is that they do it, they do it
without any human having to know that they are doing it or
having to interpret the bit patterns and tables, or even
knowing which bit patterns are used for which purposes, etc.
And in some cases the machines do it against the wishes of
human users (e.g. when the system doesn't do what the
designer or customers wanted it to do, as happens all too
often with modern operating systems.)

Other aspects of intentionality, including reference to
external entities are more complex. Some aspects require a
motivational system (e.g. not just thinking about something
but caring about it or caring whether your beliefs are true
or false). How to get some aspects of motivation into a
computer (or any other machine) is something I am working
on. Ultimately it depends on the functional architecture.

From this standpoint the assumption that there's a clear
division between things with and things without "original
intentionality" is mistaken. Neither is it just a matter of
differences of degree. There's a discrete set of
capabilities and different subsets can coexist. No subset
corresponds uniquely to any pre-theoretical notion of
understanding, or intentionality. (That claim needs
argument.) Of course, different theorists can decide to
attach the label "intentional" to some particular subset,
but that's far less interesting than trying to understand
the full range of possibilities.

So instead of asking whether X does or does not have
original intentionality we should ask "What sort of
intentionality does X have?

This seems also to be relevant to questions about
intentionality in other animals.

Aaron
[*]
Plain text copies of the paper, and a sequel called
"Reference without causal links" (ECAI 1986) are available
online:

    http://www.cs.bham.ac.uk/~axs/misc/machines.think.ijcai.85
    http://www.cs.bham.ac.uk/~axs/misc/reference.ecai.1986

Compressed postscript versions which print more compactly
are in the Birmingham "Cognition and Affect" ftp directory

    ftp://ftp.cs.bham.ac.uk/pub/groups/cog_affect/