Exploring the Lack of Dialogue in
Computer-Supported
Collaborative Learning
Mark Guzdial and Karen Carroll
With David Jaggie
Collaborative Software Lab
College of Computing/GVU
Georgia Institute of Technology
http://coweb.cc.gatech.edu/csl
Story
We have theories and mechanisms for
how people learn collaboratively: Dialogue
But they don’t fit what we measure in
actual CSCL practice
Far too little dialogue
Yet, we do measure individual learning
Where is it coming from?
Some potential answers: Vicarious
learning, mix of media and experiences
Guzdial & Carroll
2
Where Collaborative Learning
Comes From
Jeremy Roschelle’s Convergent
Conceptual Change
Heisawn Jeong’s Shared Understanding
  Learning results from and through
dialogue
Guzdial & Carroll
3
Convergent Conceptual Change
Jeremy observed cycles of:
Posing hypotheses,
Testing the hypotheses, and then
Discussing the results of the test and the
formation of the next set of hypotheses
Guzdial & Carroll
4
Example
 D: But what I don’t understand is how the lengthening,
the positioning of arrow
 C: Ooh you know what I think it is? It’s like the line. Fat
arrow is the line of where it pulls that down. Like see
how that makes this dotted line. That was the black
arrow. It pulls it.
 D: You’re saying this [dotted line] is the black arrow
 C: Yeah
 D: And I pull it the other arrow [points to vel with mouse
cursor] like
 C: like on its hinge. It pulls the other arrow on the hinge
down to the tip of the black arrow
Guzdial & Carroll
5
How much dialogue is there?
49 dialogue turns (from paper) in 30-45
minutes (Roschelle, personal
communication)
Heisawn Jeong’s results are not dissimilar
Also emphasize significant individual
contributions and turn-taking
Guzdial & Carroll
6
What we see in CSCL:
The “Dirty Little Secret”
 In voluntary usage, less than 1.0 note per student per
week, with rarely more than a single turn per thread
CaMILE and Newsgroups: Average 0.4 notes per student per
week (Guzdial, 1998)
Average length of a thread: 2.2 notes
Averaging over 400 newsgroups, 2.1 notes per thread (Terveen
and Hill, 1998)
CSILE: 2.69 notes per thread (Hewitt & Teplovs, 1999)
CoNote: 0.80 notes per student per week (Davis & Huttenlocher,
1995
AnswerGarden: 0.10 notes per student per week (Ackerman,
1994)
Guzdial & Carroll
7
But Learning IS Happening!
Three studies with similar interaction patterns,
but with externally measurable individual
learning results.
ReCoNote (Miyake & Masukawa, 2000): 0.58 notes
per student per week
SpeakEasy (Hoadley & Linn, 2000): 1.3 notes per
student per week
CoWeb (Rick, Guzdial, Carroll, Holloway-Attaway, &
Walker, 2002): 2.22 notes per student per week
Guzdial & Carroll
8
How is it happening?
What are the processes of meaning
construction (Koschmann, 2002), if not the
kinds of dialogue that Roschelle and
Jeong describe?
Guzdial & Carroll
9
Some Hypothesized Mechanisms
(From Expected to Emergent)
Vicarious learning (Lee et al., 1999;
McKendree et al., 1998): “I feel that others
are speaking for me, and I learn from
watching their dialogue.”
Dialogue served as catalyst for external
reflection.
Dialogue incited or interacted with
classroom (or other external) discussion.
Dialogue is reflecting external discussion.
Guzdial & Carroll
10
Study Design
Using log files, identified users who
Had posted to a CSCL system (CoWeb), so they
knew the mechanics,
Visited a page,
But chose not to post.
We replicated and printed the discussion as it
existed at the time of not posting.
We presented it to the student and ask them
why didn’t they post.
Guzdial & Carroll
11
Interview Guide
 It looks as if you read comments on this page, but didn't post. Can
you remember why?
 Did you ever go with something to say and find someone else had
already posted it? Did you ever come up with something to say, go
away to think about it, and return to find someone else had said it?
 When you read comments which sparked something you wanted to
say, did you generally post it right away or go away and think about
it for a while first?
 Did you ever think of something to say and then bring it up later in
class instead of posting?
 When you went to the CoWeb, did you usually go with something to
say, or were you just browsing through other people's comments?
Guzdial & Carroll
12
Two Sets of Interviews
With English students, where we had
measured learning (Carroll)
These are the quotes that are in the paper
With CS students this last semester
(Jaggie)
No evidence of learning per se, but looking
for similar mechanisms
Guzdial & Carroll
13
Evidence of Vicarious Learning
K: Apparently, you looked at this page. My impression from looking at
the CoWeb is that you posted pretty frequently. But this is one
where you didn’t post—where you looked at it and you didn’t write
anything. Do you remember this?
Student P: Yeah, yeah, I remember this.
K: Do you have any idea why?
Student P: Oh, yeah, I saw this part, and I was going to answer it, but, I
don’t know, I,… before I could answer, somebody already came up
with the answer, StudentC or StudentJ came up with the answer.
K: Did you ever find that you had a question you wanted to ask, and
then somebody else had asked it already?
Student P: A couple of times, yeah. Yeah. I guess, even if I did have a
question, I mean, the questions were there, but it just didn’t come
out, somebody had already worded it, so I’m like oh yeah, I wanted
to-I was wondering about this, too.
Guzdial & Carroll
14
More on Vicarious Learning
K: Did you ever go with an idea of something you wanted
to say and find that someone else had already posted it?
Student Z: mm-hmmm. Several times. … I probably would
have had some interesting discussions myself if I had,
but like, I like when Student P, how he, I like reading
what he writes, too, ‘cause it’s very interesting to me,
‘cause he thinks sorta like I do; like he likes to have facts
to base things up; I understand how he thinks, and like a
lot of people in the class, a few of them have the same
views and a few of them don’t and I like, it just interests
me, you know, kind of, I like how people think.
Guzdial & Carroll
15
More than Vicarious Learning:
Vicariously Resolving Conflict
David: How would you use the CoWeb to study for the midterm?
Student 3: I looked at the question, and then I try to figure out the
questions, and I looked at the responses.
David: Did the responses generally match what you came up with?
Student 3: Sometimes yes, sometimes no, depending on whether I was
right or wrong, or whether they were right or wrong. Or both.
David: What would you do it there's a discrepancy?
Student 3: Hmm...if there's a discrepancy between 2 answers? Just
wait and see what happens.
David: Just hang back and see who wins?
Student 3: Yeah.
Guzdial & Carroll
16
Learning Benefit without Posting:
Dialogue as Catalyst for Reflection
K: Somebody said what you were going to say?
Student P: Right, like, I found stuff on this, I mean I came
here, I saw this and I decided to do some research, so I
went to some other web page and I looked it up and I
found out stuff, I was actually ready to post, I come
back, and somebody had already done it…Like I used
this as kind of, you know, something that would help me
think about a particular topic and a particular direction,
like when you just read a book, it’s really difficult to pick
a subject and go in a particular direction, like if
somebody thinks about something and gives a question,
it’s easier to channel your efforts towards a particular
thing.
Guzdial & Carroll
17
Research as Dialogue
Incited by CSCL
David: Ok. Is there ever a time you saw something on the
CoWeb, and had a question about it, and then went off
and did your own thing—looked it up, did some...
Student 4: A lot of times it's faster, if it's a really dumb
question--really pretty simple, that I should have known-or that a lot of people will know, then instead of posting
it...there's always a bunch of groups working on the
same project, the same program, at any given time, so
I'll just find someone else I know and ask them if they
know it. And then, if I ask a couple people and they don't
know it, then I realize it may not be that easy of a
question, so then I'll go ahead and post it.
Guzdial & Carroll
18
Interaction between Class and
External Discussion
K: If you could just look at this and see if you can remember when you
looked at it and why you didn’t post anything.
Student Z: Trying to remember…the majority of the time, when I looked
at a page I was also trying to get other ideas from what the class
was doing. See how the class was responding to things. ‘cause
usually from that, the class would pick up on attitudes of the
teacher, and stuff like that; that way I can gear myself towards the
teacher, ‘cause every teacher you write to is different, had
completely different writing style than I did in high school, and
sometimes when I can go through, I can pick up other ideas and
other views of things that are brought up in class, ‘cause I think this
was on like the open forum, and from that I can kinda understand
what other people were thinking, and possibly get an idea of where
the teacher wants us to go with things.
Guzdial & Carroll
19
Posting in Response to a Group
(External) Discussion
David: When we were setting up this interview, you said
you didn't always post under your name. You'd post
under your group's name.
Student 4: Well, a lot of times, we're working on the group,
working on the program...the way we split up the tasks,
it's never like "One person do this, one person do this,"
it's usually "two people work on this part, two people
work on this big part." If they want to split it up between
that, then whatever. The work is usually two people...at
least two people at the same time. So when we post…I
don't know if it just felt uncomfortable, or like I'm kind of
taking the credit for the question...usually we just post
under the group name so that everybody gets equal
credit.
Guzdial & Carroll
20
Summary:
Early Hypothesis Generation
The most significant point of this paper: We
don’t know the mechanism of learning that’s
going on in actual CSCL practice.
Our studies’ seek to generate and gather
evidence for hypotheses for these mechanisms.
There is probably more than one.
We do not have evidence that any of these
mechanisms do lead to learning.
We only offer evidence that these
mechanisms/strategies are occurring.
Guzdial & Carroll
21