There are two streams in our group, those who focus on peer interaction (and are also more cognition-oriented) and those who focus on larger groups (and are more technology-oriented). The workshop is structured around this duality.

Click here to know more on the Task Force 5: Collaborative learning and the ESF programme "learning in humans and machines".

Thursday, September 21st.

9.00 	Warming up. 
	Worshop schedule, practical stuff, task force managment.

---	Session I  : Analysing and modelling peer interactions

9.30 - 10.15  	"Analyses of collaborative computer assisted learning: 
		towards a review."

		Karen Littleton, The Open University
		Kim Issroff, University College London

Abstract	In this presentation, we discuss the ways in which researchers have approached
		the analysis of collaborative CAL. We describe indices which have been used in
		interpretating collaborative interactions and some of the theoretical
		approaches that inform these analyses. We will illustrate with reference to a
		range of papers some of the approaches which have been used. We will also
		describe two small scale case studies which will illustrate, with reference to
		our own data, some of the problems confronting researchers interested in the
		analysis of collaborations. Our attempt to document the various approaches
		reveals a complex picture and points to the need for integration of the
		relevant indices in studies of computer-supported collaborative learning.

10.30 - 11.00 	Coffie-Tea break

11.00 - 11.45	"Explaining to Oneself and to Others"
         	 Rolf Ploetzner & Michael Preier (University of Freiburg, D),
          	 Pierre Dillenbourg & David Traum (Univerity of Geneva, CH)

Abstract: While the junior scientist Rolf Ploetzner visited the senior
  	  scientist Pierre Dillenbourg in his research laboratory, they
          developed together with David Traum an experimental design to
          investigate whether systematic explaining to each other leads
          to more learning than just systematic explaining to oneself.
      	  That were happy, enlightening and productive days!
          Later, being (almost) alone again, they realized that nobody
          had the time and money resources to actually conduct the research.
          Everyday work had us all back (be careful, I don't say that the
          happy days were over)!
          Fortunately, Michael Preier was in search for a topic for his
          Diplom-Thesis. After talking to him, he picked up the ideas
          developed so far. With some modifications, the study is
          currently conducted by him.

11.45 - 12.30 	"The limits of metaphors: Some discrepancies between 
		'distributed artificial intelligence' and 'distributed cognition'
		Pierre Dillenbourg (University of Geneva, CH)

Abstract: Distributed articificial intelligence attempts to improve the efficiency of 
	systems by distributing processes over different processors. It relies on 
	strong empirical bases, but still lacks of a unifying theoretical framework. At 
	the opposite, DCT renewed the interest in the socio-cultural theories. This 
	heritage provides DCT with an elaborated theoretical background, but its 
	concrete implications of DCT are still rather unclear. These respective 
	weaknesses and strengths should lead us to consider DAI and DCT as 
	complementary approaches. Moreover, both approaches offer metaphors 
	which appear to be easily transferable to the other one. Actually, one can 
	always transfer metaphors at the superficial level. However, a second 
	analysis of metaphors reveal discrepancies between the assumptions 
	carried out in DAI and DCT. The 8 discrepancies concern the 'agent' metaphor,
	the 'joint system' metaphor, the 'shared udnerstanding' metaphor,
	the 'isomorphism' postulate, the 'conflict' metaphor, the importance
	of noise, the redundancy of agents and the generalisation of group
	size. I points out these discrepancies, not to fossilise them,
	but to suggest research directions: to model the impact of social
	interactions on cognition by developing computational models in which
	dialogue operators are used within inference engines. 

12.30 - 14.30	Lunch

14.30 - 15.15  "Analysing argumentation in collaborative problem-solving dialogues"

		Michael Baker (C.N.R.S., COAST research team, Ecole Normale Superieure
		de Lyon, France)

ABSTRACT:  Despite some recent negative results concerning the role of 
	interpersonal conflict in producing cognitive change (e.g. Blaye, 1988), some
	authors	(e.g. Mavarech & Light 1990) have conjectured that the way in which such
	conflicts are resolved by argumentation in cooperative problem-solving
	dialogues may be more important than the incidence of conflict itself.
	In this presentation I attempt to take this conjecture seriously : I
	describe some interactional mechanisms and phenomena of argumentation
	sequences, and the different types of cognitive change with which they may
	be associated. After a short review of different theoretical approaches to 
	argumentation,and the associated analysis methods, I apply one such approach -
	dialogic logic - to the analysis of argumentation sequences taken from a corpus of
	dialogues produced by pairs of students solving problems in physics. 
	I then describe the expressive limits of this argumentation model, and
	describe two other important dimensions of argumentation dialogues : (1)
	the interactive dimension (negotiation of the meaning of domain knowledge
	and problem solutions), and (2) the epistemic dimension (the role that
	explicitation of different types of knowledge plays in argumentation). 
	In conclusion I describe ways in which research on analysing and modelling
	argumentation dialogues in learning sit

15.15 - 16.00	"What Machine Learning Transmutations have to do with Negotiation
        	in Teaching-Learning Dialogues"

		Engelbert Mephu-Nguifo (University of Artois, F)

Abstract:	During this talk, I would describe Machine Learning Transmutations, which
		are operators that make knowledge changes in the knowledge space. Then I would
		show the relationships between these operators and the set of "Transformation
		Functions" described in Michael Baker's paper entitled "A model for Negociation
		in Teaching-Learning Dialogues".
 		Finally an opened discussion would focus on the contribution of these operators
		in the modelization of collaborative learning.

	(I put this talk here because Engelbert said it will be closely
	connected to Michael's papers on dialogic learning)

16.30 - 17.15	 "Aspects of Conversational Agency"

		David R. Traum (Universite de Geneve, CH)

Abstract: 	Designing an agent to particpate in natural conversation
		requires more than just adapting a standard agent model to perceive and
		produce language. In particular the model must be augmented with social
		attitudes and a notion of discourse context. In this talk, I will
		focus on two such necessary extensions which are also absent in
		traditional speech act work: Obligations and Grounding. We claim that
		obligations are a necessary adjunct to beliefs, desires, and
		intentions as inputs for deliberation in social agency, and are also
		the main effect of requests. Grounding, the process of adding to the
		common ground or set of mutual beliefs between conversants, has also
		generally been oversimplified or ignored in previous speech act
		work. I briefly describe a computational theory of grounding which
		treats traditional speech acts as multi-agent acheivements which are
		completed only with participation be both the initiating speaker and
		the responder.


17.15 - 18.00 	Discussion of session I (if we are not exhausted)

Friday, September 22nd.

--- 	Session II : The relationship between situated cognition	----
	and distributed cognition (joint with members of task force 4)

9.00 - 12.30	Short contributions (20 min) from various members followed by 
	a discussion. So far, the following people said they might contribute
	on this topic by choosing some concerte case of discussion:

	Edith Ackerman (Universite de Provence, Fr)
		Title to be announced
	Bob Lewis (Universite de Lancaster, UK) 
		Title to be announced
	Ulrich Hoppe (University of Dusiburg, D)
		"What kind of cognitive and educational theories do we need
 		to support collaborative (group) learning in computerized
	Christofer Sinha (University of Aarhuis, DK)
	Pierre Dillenbourg (University of Geneva, CH)
		"The implications of situated an ddistributed cognition
		for the design of interactive learning environments".

12.30 - 14.30 	Lunch

--- 	Session III:  Collaboration in large groups	-----

14.30 - 15.15 	"Problem orientation as a method and as a 
		foundation for CSCdistanceL

		Lone Dirckinck-Holmfeld, Aalborg Unversity (Dk)

Abstract	Collaborative learning is considered a phenomenon of learning in which
		there is a mutual interdependece between individual learning and the
		development of the collaborative community. In collaborative learning at
		distance, this community should be created in "a virtual environment" in
		which a computer-based application designed for collaboration has an
		incorporated role.
		I would like do discuss how computer conferencing as an example of a
		communication tool mediates the building of "a virtual environment", as
		well as I would like to talk about our methode: Soft Dialectics.

15.15 - 16.00	"Competence development within an administrative work environment:
		Interplay between humans and software assistants"

		Carl Gustaf Jansson (KTH and University of Stockholm)

Abstract	Software assistants adaptive to people s individualized working habits for
		particular tasks are being rapidly developed. Such tasks include electronic
		mail services, information retrieval, meeting scheduling, document
		preparation and tele/video conferences. We are interested in the study of
		how an integrated set of such software assistants can be used by  a
		specific user group. In particular we are interested in the study of
		competence development, not only in individual users and software
		assistants (adaptive agents)  but in the total system of user group and
		software assistants.  Current projects and competence of the research group
		include multi agent software technology, machine learning, communication
		models, design and evaluation of adaptive interfaces in general and systems
		for help, explanation and filtering of information in particular.

16.00 - 16.30 	Tea break

16.30 - 17.15	"Intelligent support functions in distributed, group-oriented 
		learning environments - towards a taxonomy and a methodology"

		Ulrich Hoppe (University of Dusiburg, D)

Abstract 	Based on distributed computing and networking technology we face a new
		type of educational computing environments that explicitly facilitate
		group learning. It will be discussed what computerized support for group
		learning can provide beyond the mediation of human-human communication
		(synchronous CSCW) and beyond the distributed management of multimedia
		material. Given the current state of the art in cognitive modeling and
		diagnosis, student modeling will be essentially restricted to individual
		assessment. However, individual models/assessments can be used to
		intelligently inform group learning scenarios. Possible support functions
		include: balanced composition of learning groups, selection of learning
		material (exercises) for a given group, and teacher/tutor information.
		Different functions of computerized agents in group learning environments
		require different computational mechanisms. As an extrapolation of
		existing experience, we can characterize typical functions and related
		requirements and propose specific computational approaches to meet
		these requirements.

17.15 - 18.00	Discussion

Saturday, September 23rd.

---	Session IV : Merging learning in pairs and learning in large groups   ------

09.00 - 10.30 	"Theory debate: Can we use the same theories for studying collaborative
		learning within pairs and within large group?"
		Animator: Claire O'Malley

Abstract	Most cognitive theories related to collaborative learning concern
		basically a few people working closely together. When the group	
		is larger, theories rely more on social psychology, anthropology,
		and so forth. Give the recent devlopments of CSCL, there
		is a trend to generalize the findings observed in such peer interactions
		to current practices involving large groups, without face-to-face
		contact and often with a boradly defined goal. Are the peer
		interaction theories still relevant? Which theoretical development
		do we need for understanding the effects of learning in large
		groups ? 

10.30 - 11.00	Coffied break

11.00 - 12.00	"Applications debate: Can we use the same tools for supporting collaborative learning 
		within pairs and within large group?"

		Animator: Bob Lewis

Abstract	The type of support provided by computers to collaboration in large
		groups often uses a narrower bandwidth that the tools used
		to support collaboration in small groups : while small groups can share
		whiteboards and other applications, the support offered by
		larger groups is often limited to texet (e.e e-mail, WWW, MOO,
		bulletin boards, ...). This difference can be explained by
		the fact communication costs increase exponentially with the
		number of participants. Are there other intrinsic differences
		between these two types of tools?

12.30 - 14.30	Lunch break

14.30 - 16.00	Planning TF5 activities for next year.


Dr. Pierre Dillenbourg TECFA (Educational Technology Unit) Faculte de Psychologie et des Sciences de l'Education (FPSE) Universite de Geneve, Route de Drize,9, CH1227 CAROUGE SWITZERLAND Phone: (+ 41.22.) 705.96.93 Fax: (+41.22.) 342.89.24 E-mail: pdillen@divsun.unige.ch WWW: http://tecfa.unige.ch/tecfa-people/dillenbourg.html