NetNews Usenet Archive 1993 #3/NN_1993_3.iso/spool/comp/ai/philosop/7354

Internet Message Format | 1993-01-21 | 8.6 KB

Path: sparky!uunet!spool.mu.edu!agate!doc.ic.ac.uk!warwick!uknet!edcastle!cam
From: [e-mail] (Chris Malcolm)
Newsgroups: comp.ai.philosophy
Subject: Re: Colbys-Weizenbaum and Overcoming Depression 2.0
Message-ID: <[e-mail]>
Date: 21 Jan 93 12:24:44 GMT
References: <[e-mail]> <[e-mail]> <[e-mail]>
Organization: Edinburgh University
Lines: 150

In article <[e-mail]> [e-mail] (Manfred Stede) writes:

>The origin of this discussion, however, was a program aiming to be an
>artificial psychotherapist.  Someone's posting advised to boil this
>application down to the absolute minimal requirement: a tool that you
>can talk to, that munches on the words for a second, and throws some-
>thing back at you, such that some of your words reappear, and you do
>that for a while, and you achieve self-elucidation.  Well, if this is the
>purpose, and I don't care about what the program utters anyways, I might
>as well talk to my cat.  Or to myself, which may in fact be much more
>useful, because I could give myself sensible answers, couldn't I.
>No need for a self-elucidation program.

But the trouble with talking to yourself, or your cat, as anyone who
has actually seriously tried this knows only too well, is that you (or
your cat) don't ask yourself unexpected questions. That is why ancient
systems of psychotherapy make such use of random oracles (reading
bones, cards, tea-leaves, etc.). And that is the reason why puzzle
solvers who are stuck are advised to try "silly" ideas, and to try
random mutations of suggestive material, in an effort to jolt the mind
out of its rut.  It's the old generate-&-test paradigm. The human mind
is a lot better at testing than generating, however, hence the
popularity of artificial aids in the generation part of the problem
solving process.  That's why a silly clueless word-crunching program
can be a useful adjunct to self-elucidation. And silly and clueless as
it may be, it would be hard for it not be better than opening the
Bible (or whatever) at random, for the utility of which there is great
deal of evidence.

Your kind of criticism is based on the presumption that _if_ a
psychotherapeutic program is any good, then it _must_ possess, within
itself, certain psychotherapeutic qualities. An obvious one of these,
which the program can't avoid lacking, is understanding the distressed
person. But a therapeutic program which does its work by engaging in a
dialogue with a person is a participant in a process involving a very
capable mind -- the mind of the human user. It can help that mind if
it can provide facilities which that mind can find helpful. These can
be very simple facilities indeed. For example, because I have a poor
memory, I find pencil and paper very helpful in solving problems of
all kinds. Including difficult problems involving sexual relationships
with others.  But there is absolutely _no_ problem solving ability in
the paper and pencil I use. For a start they know nothing about sex.
Yet I find them invaluable aids.

>If the standards remain a little higher and we truly want an advice-
>giving system, then one of my objections is that with the current state
>of NLP it just won't work and would not be of any use: There is no
>sublanguage to be pinned down here, to which you could connect your KB
>such that you get some inferences right and come back with useful
>advice.  The program designer cannot anticipate what the problem of the
>patient is, what s/he will be talking about, and how s/he will be talking
>about it.  And the user is bound to notice at some point how little that
>dialogue partner actually grasps; frustration is a likely reaction.

It is very easy to use simple ELIZA-like techniques to create a
program capable of imparting useful information in the form of a
discussion, or a question and answer session. Of course the program
often makes mistakes, misunderstanding the user and making a
completely irrelevant response. But people so that too. We respond by
rephrasing our questions. Sometimes we have to do this a number of
times. It works with ELIZA-ish programs too. In fact, it is much
easier to rephrase your questions to get the answer you want,
precisely because there is no (mis)understanding involved. Just as a
hyper-card system can help you to learn something better than a plain
book, so an ELIZA-like system can do a bit better than a hyper-card
system. In this sort of application the user doesn't get frustrated by
noticing the mechanics of the system; the user learns how to use the
way the system works to get what she wants out of it.

Of course if a user thought the program was really _understanding_
their problems, then frustration and tears are inevitable. But this is
simply due to misunderstanding by the user. The solution is education
-- which can in this case be easily provided by the supplier or the
machine.

>If the sublanguage cannot be restricted due to the domain (as it is
>the case with the psychotherapist), it becomes quite difficult, if not
>impossible, to explain to someone with no linguistics background what
>the program can handle and what not.

You don't need a linguistics background to understand how an
ELIZA-like system works! And adding simple parsing of logical
relations between clauses is easy to implement, and to understand as a
user. I once built an ELIZA-like system whose mission in life was to
explain how it worked, and to engage in AI/philosophy of mind debates.
School kids played with it for hours, enjoyed discovering how it
worked, and had no trouble understanding it. Once they had grasped
its capabilities, a favourite trick was to engage in a carefully
staged dialogue which would convince a visiting teacher or parent that
it really understood! (It was particularly good at Chinese Room
arguments :-)

The program doesn't have to make a good response every time. Irrelevant
responses are quite acceptable provided there are ways the user can
work round to getting the required response. A program which interacts
with a _very_ sophisticated language user can employ the abilities of
the user.

>No reason to get AI involved in this business, no need to advertise
>anything as a computer program that analyzes your psychatric problems
>for you and helps you solving them.

This is your misconception. You think that if AI is involved there
must be some element of "real" analysis and understanding involved.
But as I have argued, an _aid_ to self-elucidation need not duplicate
human cognitive capabilities, since it can borrow them; it only needs
to be able to offer useful facilities which the user can employ.  You
would then like to say "Ah, but in this case there is no AI, it's just
a program, an animated book." What you don't realise that this kind of
animated book can usefully employ the techniques of AI to better do
its job of being an animated book. AI can be (and is routinely) used
in applications without a trace of, or the slightest intention to
have, "real" understanding. That's a research goal about which there
is a lot of debate :-) It has nothing to do with the business of
building useful interactive books or self-elucidation aids, since
useful versions of these can be built without "real" understanding.

Of course they would be even better with "real" understanding. But,
like today's calculators, they can do a pretty useful job without it.

It interests me that Weizenbaum's secretary was smart enouh to realise
the therapeutic possibilities of ELIZA, whereas Weizenbaum was
outraged by what he mistakenly considered to be the philosophical
implications of what she was doing. As a general rule women find it
much easier than men to use oracles like Tarot-reading gypsies,
horoscopes, I Ching, etc., to help them elucidate their problems,
while men are much more likley to be outraged by what they
(mistakenly) think are the philosophical implications of asking a pack
of cards a question.

In psychotherapeutic programs have we discovered a use of computers
which women will take to quite naturally, but which men will have
great conceptual difficulties with?

(There's an intersting philosophy of mind issue behind all this.
Manfred's view depends on seeing intelligence etc. as being primarily
a property of an agent, whereas I see it as primarily a property of
the interactive process between an agent and something else. I no more
expect to find intelligence _inside_ myself than I expect to find it
in the Chinese Room :-)






--
Chris Malcolm    [e-mail]          [phone]
Department of Artificial Intelligence,    Edinburgh University
5 Forrest Hill, Edinburgh, EH1 2QL, UK                DoD #205