>From hcl@cs.pitt.edu Mon Apr 1 13:07:22 2002 Baker: A growing number of introductory programming courses have moved to pair programming (a version of extreme programming with n=3D2) for assignments. Students work on projects in pairs, alternating the roles of "driver" (who sits at the keyboard) and navigator (who sits nearby). Typically, the navigator's job is to direct while the driver's job is to type, compile, etc. Planning, debugging, etc. are therefore all done collaboratively. Instructors who have adopted=20 this approach report significant gains in both scores and interest, and formal evaluations have been done to back it up. The more that I think about, the more I am convinced that pair programming involves pretty much non-stop negotiation ("should we try X? why is that error coming up? etc.). Is it possible that the gains from pair programming are due to the necessarily high levels of negotiation?=20=20 What exactly is the relationship between negotiation and collaboration? Baker talks some about cooperation and collaboration, but I'm still not seeing the larger relationships between the ideas. =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D Lund, Baker, Baron: Maybe it was due to space constraints, but I was disappointed in their proposed methods of guidance (p.212). The first is just a traditional overlay approach. The third is very vague - detecting the discrepencies is easy, but the truly interesting idea is to let the computer act as mediator when the students can't agree, I'd like to know how this will work. Other than #1, aren't there more fundamental approaches that=20 should be explored first? Like simple hinting based on the domain rules, or mal-rules (or bugs) to identify common mistakes? Something fundamental to #3 would be to explore common points of contention which would be=20 helpful in knowing how to resolve them.=20=20 McManus & Aiken: I know it is cheap to complain about small subject pools, but they only had 9, and all were cs/math types more prone to be comfortable in such an environment. Given that, I don't know if I'd agree that 3.0/5 is necessarily a good result, especially given the subjects (that's a "D" in most classes). NOTE: The sample dialogues are very similar to those I have heard=20 take place in pair programming. =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D >From rwilson+@pitt.edu Mon Apr 1 13:07:22 2002 Baker (1994) A great paper: it provides a broad conceptual framework for discussing negotiation and a relatively precise set of defining criteria and a model of refinement. This work might provide the basis for a computational realization of many of the ideas about education that have been advanced by critical theorists in education. The distinction between belief and acceptance is especially important for constructing an adequate student model, as Baker notes on p. 248. I take issue, however, with Baker's acceptance [? :-)] of context as "the union of epistemic states of dialogue agents" (p. 224): this is conceptually insufficient (though perhaps empirically adequate) because it seems to imply that socio-cultural factors are either of secondary interest or can be captured by epistemic states. Lund, et. al (1996) C-CHENE is an interesting attempt to flexibly structure collaborative interaction between humans and allow for the collection of corpora that avoids some of the complexity and ambiguity of unconstrained natural language dialogue. Perhaps because of the domain, the authors do not draw upon the distinction between belief and acceptance. Work on the robustness of misconceptions and the lack of transfer of "school knowledge" is suggested by the following phrase, offered by a critical pedagogue in cconn= ection with university physics: "Tell us what you want to hear and we'll tell you what we think." McManus and Aiken (1995) An interesting approach. Too bad that the study had no measures of cognitive gain in a domain other than cooperative skills. On pages 315-16 the authors state "Interpersonal and small group skills are achieved as the students discuss their work ...". Per the recommendations of Chin, I don't see that the authors have provided sufficcient evidence to support their claim. Assessing the system without using objective measures (a la PARADISE) seems rather weak (but, hey, who am I to criticize?). Kneser and Pilkington (2001) The fact that the authors draw on the Birmingham School of discourse analysis, plus their reference (with Baker) to the differences in the distribution of dialogue roles (p. 64), suggests to me that they too may be doing work that is compatible with critical theory in education. However interesting their work from a qualitative view, it suffers from some issues of validity (Chin) and credibility. First, on page 73, the authors note that group scores seemed to vary considerably (how consideraably?) depending on whether the student were full or part-time. Doesn't the lack of a quantitaive analysis of this variability undermine the generalizability of the work? P.S. For a "blurb" on Critical Discourse Analysis, see http://www.hum.uva.nl/~teun/cda.htm (which I just found today). =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D >From mbell@cs.pitt.edu Mon Apr 1 13:07:22 2002 Monitoring Computer-Based Collaborative Problem Solving: I don't want to sound anti-technology, but the tool being presented in McManus's paper is something I am highly skeptical of in terms of its ultimate value. To show why, consider the calculator: After its widespread use in high-school and college, the ability to do math mentally dropped. I know this is true of myself and others. Now consider the same sort of technology drawn to cover not something valuable but non-essential, such as arithmetic, but something absolutely essential such as basic communication skills, politeness, etc. Is not that exactly what McManus and Aiken are presenting? Their tool is essentially a conversational calculator, a system that tells you when to say and not to say "please", "thank you", "I think that...", etc. The import of wide-spread use of such technology is something that, frankly, I find disturbing. Assume that this did not visibly diminish independent thinking and social skills, for the sake of argument. All the same, has it truly encouraged people to integrate people with each other, or is the computer here essentially being used to seperate people from each other, to narrow the communication chanel? The narrowing, granted, is designed to exclude rudeness and conversationally "inappropriate" rhetoric, but isn't such exposure a necessary part of human development, not simply of collaborative skills but also relational skills? I fear I could be seen as reactionary and anti-technology here, but really the implications of this "tool" frighten me. What would be the long term impact of wide spread use of McManus's technology? Would people lose either the ability or the desire to work with each other apart from cognitive and communicative prostheses? Are we simply offloading our social and linguistic skills onto a machine so that we no longer have to put up with the "messy" task of relating to other people? -------------------------------------------------------- Modelling Dialogue and Beliefs as a Basis for Generating Guidance in a CSCL Environment: Ditto the above concerns, although this paper doesn't bother me as much, perhaps because the seem more interested in using the technology for _learning_ about people, rather than in _enforcing_ the system's preprogrammed conventions upon them. That said, does their claim that forming a descriptive grammar of discourse is impossible but a normative one possible a bit suspect? What is "normative" here? If a general grammar of discourse is impossible, why would a normative one be possible? How do they know there is such a thing as a normative discourse? It seems what they're really doing here is radically constricting the communication channel, and then monitering human linguistic behavior within that constrained channel. If they are doing the latter, the results seem potentially intersting, although these were not discussed in detail so it doesn't seem clear what they might have learned. If they really can successfully monitor shifts in beliefs and changes in belief spaces, that would be very useful in building knowledge of grounding. --------------------------------------------------------- The Tutor's Role: Their use of ESA to distinguish between different student roles was fascinating, although it seems most of their reliable conclusions they drew regarding the tutor. These seem to match simple common sense: namely, that the tutor is the "powerful" interlocuter in a tutorial dialogues. This seems a little disappointing, as having the research show what is already widely accepted doesn't seem to be a very valuable contribution of the technique, hence does not well-illustrate its strengths. The ability to differentiate, even potentially, between different sorts of initiation and different uses of social power in discourse is really very interesting, though. =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D >From alandale@cs.pitt.edu Mon Apr 1 13:07:22 2002 Baker What "complex issues" does the argue strategy involve? How could a system make use of this distinction between beliefs and acceptance? Do any existing systems (tutoring or other) do this? =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D >From roque+@pitt.edu Mon Apr 1 13:07:22 2002 "A Model for Negotiation in Teaching-Learning Dialogues" says negotiation is relatively new to ITSes, and this is "due in part to constraints of formalisability imposed by AI programs, which precluded until recently the implementation of genuinely mixed-initiative interaction mechanisms." This may be a trivial question, but what exactly are the contraints he's talking about? =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D >From funkydung@pobox.com Mon Apr 1 13:07:22 2002 Baker Finally - negotiation and argumentation as collaboration. It's about=20 bleeding time. Anyhow, I only skimmed this beast, but I liked its=20 premise. I only wish I'd had more time, so I could find stuff I disagree= =20 with. I'm sure my argumentation on the issue would be productive for my=20 learning of the concepts. ;) On first glance, the model he uses seems a=20 bit limited due to the small vocabulary (accept, reject, offer, etc). Kneser, et al Very interesting. I especially liked the mention of Distance Learning and= =20 MOOs. I wonder if this scheme could be used to determine student groupings= =20 that are more conducive to learning (eg 1 critic, 1 narrator, 1 intern, 1= =20 chimp, and a partridge in a pear tree) Lund, et al Delightfully short and rather unimpressive compared to ESA. I think my=20 beef with Baker is the limited vocabulary, as mentioned in my Baker=20 review. Why have a separate grammar for negotiation outside of a general= =20 dialog tagging system? McManus and Aiken The concept seems interesting and merits further research, but the=20 implementation (scripts and production rules) strikes me as "yesterday's ne= ws". =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D >From andyg@cs.pitt.edu Mon Apr 1 13:07:22 2002 What is an Artificial Learning Companion? Andy Gaydos =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D >From mrotaru@cs.pitt.edu Mon Apr 1 13:07:22 2002 M. Baker paper: I am still not convinced of the role of negotiation in learning. The author says in page 209 that incomplete knowledge claim can be used when modeling how to teach and when modeling knowledge of individual students' knowledge. I agree with this point, but how much gain comes from employing negotiation in such cases? For me, the first dialog (fig1 page 205) seems a little bit unnatural. McMannus and Aiken paper: Very interesting, no errors in student conversations (they did not went in an error state of the finite state machine)! The authors claim (page 331) that this is due to limited time for experiment or limited depth discussions. I wonder!?! Maybe people already have that finite state machine in their discussion model and there is no need to learn it. In this case, what is the use of their system? Just a statistic tool to collect skill usage??? =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D >From twilson@cs.pitt.edu Mon Apr 1 13:07:22 2002 McManus and Aiken - "Monitoring Computer-Based Collaborative Problem Solving It definitely seems from the authors' description that they are trying to move beyond a collaborative working environment to something more like a tutoring system for collaboration skills.=20=20 However, the one experiment that they showed did not support that=20 their system accomplished this goal. In fact, based on the user=20 feedback and the fact that the Group Leader was not called on even=20 once to generate feedback during the experiment, doesn't it seems=20 that the authors have only succeed in developing a realy nice=20 collaborative learning/work environment? Kneser, Pilkington, Treasure-Jones - "The Tutor's Role" How does the ESA coding scheme compare to adjacency pairs? Could adjacency pairs be used in the same way as ESA to capture who is holding the initiative? =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D >From steffi+@pitt.edu Mon Apr 1 13:07:22 2002 Baker: How does negotiation in collaborative dialogs relate to=20 self-explanation? It seems that when constructing an argument, the=20 student is forced to self-explain more, compared to solving a problem by=20 himself. McManus & Aiken: How much gets lost when one focusses on the openers, rather than one the=20 entire utterance and its contents? Students (in particular younger=20 students in their "rowdy" years, to a much lesser degree the population=20 they had in the pilot study) probably do not adhere to Grice's maxims=20 very well, and one can also expect that their utterance may not always=20 correspond to the opener - how much would these two problems limit the=20 utility of their approach? ------=_NextPart_000_001C_01C1D97F.5CF15B80--