Ruddy Lelouche (lelouche@ift.ulaval.ca)
Tue, 8 Sep 1998 13:21:24 -0400
>Received: from mail.gmd.de (postix) by zeus.gmd.de with SMTP id AA18690 (5.67b8/IDA-1.5 for <ifetsarc@zeus.gmd.de>); Thu, 1 Oct 1998 10:38:14 +0200
>Received: from mail.gmd.de (postix) by zeus.gmd.de with SMTP id AA06560 (5.67b8/IDA-1.5 for <rashev@zeus.gmd.de>); Tue, 8 Sep 1998 20:21:21 +0200
Date: Tue, 8 Sep 1998 13:21:24 -0400 From: Ruddy Lelouche <lelouche@ift.ulaval.ca> Subject: [ifets] Reply to J. Eklund: What is an ITS?
Forum website: http://ifets.gmd.de/
Email address for sending message to the list ifets@gmd.de
Web-based discussion board: http://ifets.gmd.de/discuss/
Forum's contact person ifets-info@gmd.de
---------------------------------------------------------
Hello John Eklund, and hello Victoria Satta!
John, thank you for your reaction to my ""tentative definition"": that is
how things get clarified (they were clear to me, since I was refering to my
paper:-).
>>What is important is that the system exhibits:
>>- a good "expertise" of the learning domain,
>>- and a capability of good explanations in that domain,
>> in particular a capability to answer students' questions.
>>
>>Would you agree (especially the moderators) that these two characteristics
>>could define an ITS answering Vilnis Vulfs's question for the forum's
>>purposes?
>
>
>Well... no, I wouldn't agree. An ITS, to be so called, needs to have:
>
>1. A domain model - A means of representing domain knowledge
I quite agree with that: it is the core of first generation ICAI systems.
That model gives them the good "expertise" of the learning domain, which I
was refering to (I spoke about the outcome, and you spoke about the means
to get it, but we speak of the same thing).
>2. A student model - A way of recording information about individual
>student progress in that domain. It's usually an overlay of 1.
For practical purposes, that can be derived quickly from the current
"dialogue" between the student and the system (or the tutor). By the way,
that is how most human tutors operate, since I don't know of any student
who carries his "student model" along with him/her in order to get valuable
feedback. Naturally, I refer here to tutoring, i.e. to an attempt to
transmit knowledge to the student, and not to evaluating him/her, i.e.
(schematically) assessing what parts of the "ideal" or "complete" domain
knowledge are (or are not) overlaid by the student knowledge.
>3. A tutoring strategy, which utilises 1 & 2 to individualise instruction
Here again, tutoring strategies can be used without any student model. That
is the way all teachers operate, especially in front of a multi-student
classroom. Obviously, even if there were student models available, they
could not take into account every individual student model! And good
teachers do have tutoring strategies, based on their previous experience,
on the way the student(s) react to each stimulus (in the broadst sense), on
a possible hierarchy of the domain knowledge in order to set up a good
progression in the course material, etc. There even exist attemps to model
domain-independent tutoring strategies, e.g. Collins's work.
>4. And, importantly, use an 'intelligent technology' such as an expert
>system, which is why we don't use the term 'intelligent' when talking about
>adaptive systems.
Quite agreed (although expert systems, to me, are not a technology, but a
type of approach - involving methods and techniques - of representing some
aspects of the world realities)! So here again we are on the same boat!
Indeed, that approach is the keystone of 1st generation ICAI systems, or
ICAI systems expert in their domain. It is inferencing capabilities on the
learning domain that can give them the second aspect I was initially
refering to, i.e. the capability of good explanations in the domain, and in
particular a capability to answer students' questions.
Note that I am defending the absence of a student model, not because I am
against student models, but to show that good systems can exist without
such models (of course, you can always call "student model" student-related
knowledge derived from the current session exchanges :-). In my survey
paper, I claim that tutoring strategies (your item 3) is what makes an ICAI
system (i.e. exhibiting at least my two characteristics, or your two items
1 and 4) a second generation ICAI system, or ITS. But that is all!. Indeed,
some researchers even argue that building a deep student model (i.e. a.k.o.
your item 2), going beyond the simple overlay model (e.g. taking into
account learning strategies, etc.), may not yield a very return on
investment... I must admit that I am quite close to thinking along the same
lines...
I hope that things are now clearer... We are not so far apart, after all ;-)
Best regards,
Ruddy Lelouche Phone: (+1-418) 656 2131 (X 2597)
Departement d'Informatique Messages: (+1-418) 656 7979
Universite Laval Fax: (+1-418) 656 2324
Quebec G1K 7P4, CANADA Mailto:lelouche@ift.ulaval.ca
This archive was generated by hypermail 2.0b3 on Thu 01 Oct 1998 - 11:12:15 MET DST