NetNews Usenet Archive 1992 #27/NN_1992_27.iso/spool/comp/ai/4368

Internet Message Format | 1992-11-21 | 8.0 KB

Path: sparky!uunet!zaphod.mps.ohio-state.edu!usc!usc!not-for-mail
From: [e-mail] (John Kim)
Newsgroups: comp.ai
Subject: [Summary] Self-expanding expert systems & 3 cognitive problem solving l
Date: 22 Nov 1992 01:09:48 -0800
Organization: University of Southern California, Los Angeles, CA
Lines: 218
Message-ID: <[e-mail]>
NNTP-Posting-Host: mensa.usc.edu
Summary: summary of responses to my past posting
Keywords: summary

Sometime ago I posted two requests on comp.ai for references to

1. Qualitative Reasoning based self-expanding knowledge based systems
2. literature that supports my conjecture for existence of at least three
   cognitive levels of human problem solving.

And here is the summary. (I apologize for the late response. I got married :->
and was overseas for a while, got sick, etc. :-< )

I got only one response (indirectly related) to #1 and several responses to
#2, which are as below. The first reply is to #1, the second message is my
original posting for #2 and the following ones are replies to #2.


At the moment, I'm taking a look into knowledge compilation (KC) literature
as KC seems to be keenly related to knowldege representations at different
knowledge levels, where the mapping function between adjacent levels seems to
be KC. If you are interested as I am in this area (ie, KC), a good place to
start is the April 1991 issue of IEEE Expert magazine, pp. 71-93. This is a
series of 7 articles about KC. The extensive references at the end is also
informative. Another place to look into is the Proceedings of Workshop on KC
(edited by Tom Dietterich, a Tech Report of Dept. of CS, Oregon State.
Univ., Corvallis, Oregon, 1986).

If you have any interesting results or pointers to others' work related to #1
and #2 above, or to KC, please let me know by email.

Thank you.

John

Joung-woo John Kim                [e-mail]
Computer Science Dept.
University of Southern California, Los Angeles, CA 90089-0781



~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

[A reply to #1]
To: [e-mail]
Subject: Re: (Q) Self-expanding QR-based Knowledge based systems
Date: Mon, 06 Apr 92 11:27:20 +0100
From: Michael Luck <[e-mail]>

John,
I'm not entirely sure what you mean, but there is a growing body of research
on scientific discovery, theory revision and knowledge base refinement.

booktitle = {Computational Models of Scientific Discovery and Theory
Formation},
editor = {Shrager, J. and Langley, P.},
publisher = {Morgan Kaufmann},
address = {San Mateo, Calif.},
year = {1990}}

@book{ginsberg88,
author = {Ginsberg, A.},
title = {Automatic Refinement of Expert System Knowledge Bases},
publisher = {Pitman Press and Morgan Kaufmann},
address = {London and San Mateo, CA.},
year = {1988}}

@inproceedings{ginsberg89,
author = {Ginsberg, A.},
title = {Knowledge Base Refinement and Theory Revision},
booktitle = {Proceedings of the Sixth International Conference on Machine
Learning},
address = {Ithaca, NY.},
year = {1989}}

The book edited by shrager and langley has a collection of papers that are
concerned with various aspects of computational discovery, and is probably
the best (if not the only) book available that offers a variety of research.
Ginsberg's work is concerned with knowledge base refinement
(and also (I think) with reducing a theory from one form to another). The last
paper here is interesting because it relates knowledge base refinement to
scientific ttheory revision, which seems to be relevant to you.

I'd also be interested in knowing about other responses, and about your own
interests too.

Mike.

Mike Luck                email: [e-mail]
Department of Computer Science
University College London
Gower Street
London WC1E 6BT
United Kingdom

THE SUMMARY


[My request for #2]
Your article/letter follows:
Newsgroups: comp.ai,comp.ai.edu,comp.ai.nlang-know-rep,sci.psychology
Subject: (Q) Ref on cognitive levels of problem solving
Summary: Looking for ref's on cognitive levels of problem solving
Keywords: cognitive levels, problem solving


      Could someone kindly point me to references (in cognitive
psychology, artificial intelligence, etc.) that talk about cognitive levels
of problem solving? I may not be labeling the area of my question really
well, but I'm looking for references that will support the following
conjecture:

There are at least three levels of human problem solving. I will call them
associational, causal and "hypo-causal" levels. Roughly, the A-level is where
rule-based (i.e., "compiled" knowledge) reasoning happens, the C-Level is
where to resort to when A-level reasoning fails to give an answer and where
causal reasoning happens such as in these days' qualitative reasoning
programs, and the HC-Level is where you resort to when even C-level reasoning
fails to give an answer and where something "mysterious" happens.


      I hope you get a rough sketch of my questioning context from above
brief description.

      Please don't point me to Rosenbloom et al.'s Soar material and J. R.
Anderson's work, as I already accessed them.

Thanks.

John

[Replies to #2]
Date: Mon, 7 Sep 1992 16:02:33 -0500
From: "Patricia M. Jones" <[e-mail]>
Message-Id: <[e-mail]>
To: [e-mail]
Subject: Re: (Q) Ref on cognitive levels of problem solving
Newsgroups: sci.cognitive
References: <[e-mail]>

Check out Jens Rasmussen's (1986) book on
Information Processing and Human-machine Interaction:
An approach to cognitive engineering.

His decision ladder model distinguishes between
skill-, rule- and knowledge-based cognitive behaviors.


--
Patricia M. Jones
Dept. of M&IE, Univ of Illinois,
1206 W. Green St., Urbana IL 61801
VOICE: 217-333-3938 EMAIL: [e-mail]


Date: Fri, 4 Sep 92 17:45:56 CDT
To: [e-mail]
From: [e-mail]
Subject: Levels of reasoning (or something)

As a candidate for your "mysterious" level, you should consider case-based
reasoning. Refer to "Explanation Patterns" and/or "Dynamic Memory" both by
Roger Schank for the general picture.  For more detail on current research,
see if you can get any of the proceedings from the case-based reasoning
workshops.  There have been CBR sections at both AAAI and Cognitive Science
conferences for several years now, so those proceedings would also be good
places to start.


robin
=:-{)

institute for learned scientists
northwestern university


From: N D Silver <[e-mail]>
Date: Sat, 5 Sep 92 13:22:35 +0100
Message-Id: <[e-mail]>
To: [e-mail]
Subject: Re: (Q) Ref on cognitive levels of problem solving

In article <[e-mail]> you write:
>      but I'm looking for references that will support the following
>conjecture:
> [...conjecture deleted...]

As your investigations expand into relted fields, you might find the
following reference provides some helpful evidence:

G. Altmann, The Levels of Linguistic Investigation, in Theoretical
Linguistics, Vol.14 (1987), pp. 227--239

The abstract reads:

    According to the aspect and the aim of syudy we can discover three
    levels of linguistic investigation:
    1) the homogeneity level concerned with rules and yielding
    descriptions,
    2) the heterogeneity level concerned with variability and change and
    yielding descriptions and classifications, and
    3) the process level concerned with processes and self-regulatations
    of language and yielding laws, theories and explanations.

>P.S. Please reply to me by email, since I don't regularly read this
>newsgroup. I'll post the summary if appropriate.

I would be most grateful if you would post a summary.

Thanks,
--
Nik Silver                          "Now, give me the gatekey."
[e-mail]                   "What gatekey?"
                                    "Fezzick, tear off his arms."
                                      "Oh, you mean this gatekey."
                                                   --- from The Princess Bride

--
Joung-woo John Kim                [e-mail]
Computer Science Dept.
University of Southern California, Los Angeles, CA 90089-0781