Computer Based Learning Unit, Leeds University
Leeds LS2 9JT, England, UK
The position taken here is that the notion of collaboration is better seen as a set of possible relationships between participants. The term collaboration then becomes a generic concept that has to be instantiated to define the desired relationship between participants. The goal of this paper is to briefly sketch research into different collaborative relationships that illuminates four key issues.
Examples of ongoing research are used to illustrate these issues. These examples refer to work on: a computational model of collaboration intended to support learning to collaborate [Burton et al, 1997], the use of collaboration within virtual reality environments [Brna, 1998,Romano et al, 1998], the use of collaborative student modelling [Dimitrova et al, to appear], and the notion of collaborative assessment [Bull & Brna, 1997].
There are many uses of the term collaboration in the current educational and computing literature. The Computer Support for Collaborative Learning (CSCL) field provides some indication of the diversity of the notion, and Dillenbourg et al provide a theoretical discussion of collaboration which, to an extent, glosses over this diversity for the sake of obtaining an overview of research in the general area [Dillenbourg et al, 1994].
The position is taken here that though there are many different meanings associated with the term collaboration, that these meanings are often associated with quite different educational goals. Significant issues include:
The first of the above points is used by Roschelle and Teasley to differentiate cooperation from collaboration.
``Cooperative work is accomplished by the division of labour among participants, as an activity where each person is responsible for a portion of the problem solving...'', whereas collaboration involves the ``... mutual engagement of participants in a coordinated effort to solve the problem together.'' [Roschelle & Teasley, 1995].
Roschelle and Teasley, go on to say, collaboration involves a ``coordinated, synchronous activity that is the result of a continued attempt to construct and maintain a shared conception of a problem''.
Roschelle and Teasley's description of cooperation is not really a very satisfactory definition: it is better seen as a characterising description. The problem is that even within the closest collaboration possible (under their definition of collaboration) the participants will do some of the subtasks on their own (in their heads, on scraps of paper etc).
For example, in a pilot study, pairs of students were asked to work together within a collaborative virtual environment in order to perform a `shared task' [Brna, 1998]. The participants need to divide their time between acting in the world (moving around, examining objects) and talking with their partner. This highlights the inadequacy of taking Roschelle and Teasley's description of cooperation and collaboration as a definition since in reality participants are always working on their own (making sense of things, performing calculations, moving around, writing etc).
Following on from this, and touching upon the second point above, some would argue that cooperation and collaboration are mutually exclusive. For instance, if a task is divided, and different participants work on each part of the task then during this period the participants are cooperating and not collaborating. However, the position taken here is that collaboration has to be considered both as a state and as a process. In this view, participants could cooperate in a process while maintaining a collaboration as a state. (An analogy would be that being a writer is a designation of a state but actually sitting down and writing is a necessary process that is only a part of being a writer.)
If it is accepted that people can enter into a collaborative state then there is a set of questions relating to how this state is established, maintained and exited. These issues can too easily be glossed over in classroom contexts where children are often expected to enter the collaborative state on the orders given by the teacher (i.e. the formation of arbitrary peer groups).
Knowledge about how to provide on-line support for the maintenance of the collaborative state is relatively scarce. In a study of pairs of participants asked to collaborate in surviving a dangerous virtual world, Romano found that occasionally the collaborative relationship broke down [Romano et al, 1998]. Since the pairs of participants were chosen on the basis that they had worked together successfully in other contexts, it would have been reasonable to expect them to have a good chance of maintaining the collaboration. However, the demands of the environment coupled with the role adopted by the more experienced partner led in one instance to `bossy' behaviour which fractured the sense of partnership for the participant being `bossed about'. It may be worth speculating about how any computational environment can provide support for recovery from such social conflicts.
The third point relates to the approach taken by Burton and his colleagues [Burton et al, 1997,Brna & Burton, 1997]. He takes the position that within the collaborative state there can be processes that are cooperative or even ones often associated with argumentation. Concentration on the notion of the domain-based ``goals'' of the collaboration, as in Roschelle and Teasley's definition above, has lead researchers to ignore some interesting forms of collaboration. For instance, it is to be expected that high quality collaboration may well include argumentation, and that this argumentation requires that different participants adopt, and attempt to maintain distinctly different positions. It is hard, but not impossible, to see this as ``a coordinated, synchronous activity that is the result of a continued attempt to construct and maintain a shared conception of a problem'' [Roschelle & Teasley, 1995]. It is easier to see such activities (argument, negotiation) within a framework of different active processes and roles.
The suggestion here is that while there are many different notions of collaboration, most of them are focused on collaborating to learn and not on learning to collaborate; and that they are mostly, but not exclusively, organised around the domain-based goals. That is, the primary purpose of the collaboration is on goals and the desired, shared, outcomes. If learning to collaborate is the primary educational goal then it is necessary to look at research conducted in other contexts more directly focused on how students become proficient at collaboration [Burton et al, 1997].
Burton's Clarissa (Collaborative Learning As Realised In Simple Simulated Agents) incorporates a model of multi-agent collaborative dialogue [Burton & Brna, 1996]. Clarissa was developed to simulate different ways in which agents collaborate during problem solving and learning. The emphasis is on generating collaborative discourse broadly similar to that generated by actual groups of students. The educational priority is on learning to collaborate and not just on collaborating to learn.
The fourth point is whether or not there is a formal contract between the participants. In typical classroom contexts, it is now quite common in the UK to find the notion of a learning contract --- outlining the contractual obligations between school, student and parents. Even in schools which do not make such a contract explicit there is an implicit formal contract that students will attend the school to learn their lessons, to behave in a reasonable way etc. In a collaborative partnership, any formal contractual obligations are supplemented by a set of implicit obligations (e.g. the Gricean maxims [Grice, 1975]). For example, it is implicitly assumed that a group of peers collaborating on producing a poster will not hit each other! (Whether these obligations are respected by all parties can be very problematic).
The issue of collaborative student modelling is one for which the notion of a contract is relevant. Dimitrova is extending Bull's work [Bull, 1997] on collaborative student modelling [Dimitrova et al, to appear]. In this work, the system acts as a collaborative partner to the student as he/she seeks to understand specific concepts represented by conceptual graphs presented in diagram form. A formal contract is relatively easy to construct for the system but not quite so easy to determine for the human partner.
In human-human collaboration, all parties have their own personal goals (at the very least, it is to be hoped, to learn something for themselves). The same can be true for human-system collaboration. This is so for the kind of collaborative student modelling described by Bull and being considered by Dimitrova. The system and the student are supposed to collaborate in trying to form a mutually acceptable model of what the student knows. However, both parties can `agree to differ'.
In another piece of research related to the notion of collaborative assessment, Bull and Brna examined how two students can use PeerSM, a system designed to promote learner reflection by providing an inspectable student model constructed from self evaluation, peer assessment and system assessment [Bull & Brna, 1997]. The contract between the two collaborators includes a formal obligation to provide a critique of the other collaborator's work and then discuss disparities between their own work and the work of the other participant. The two participants can be regarded as in a collaborative state entailing each of them to:
Bull and Brna report that, in using PeerSM themselves, they were both aware that further discussion was needed to finish the mutually agreed task [Bull & Brna, 1997]. This awareness could be `written into' the formal collaborative contract so that it becomes acceptable for a task to be terminated even when some unresolved issues have been acknowledged by the participants.
There are situations in which no formal contract --- implicit or explicit --- exists between collaborating partners. This suggests a form of opportunistic collaboration which is maintained by a set of implicit obligations or beliefs such as that each participant believes that the other participant can make a significant contribution to the solution of the current task, or that it is polite to answer questions in an informative manner. This is close to the concept of cooperative dialogue which underlies many approaches taken by the computational linguistics community to the issue of generating collaborative dialogue.
This brief overview of research into different aspects of collaboration has skimmed over four of the key issues that need to be considered. This has implicitly defined a space of possible models of collaboration which could be formalised (to an extent) quite quickly. The position argued for here is that: division of labour can be part of collaboration; collaboration is both a state and a process (and hence care is needed to distinguish between the two); the educational goal for which collaboration is designed has a significant effect on the nature of the collaboration; and that it is valauble to consider there being a contract between participants which has both implicit and explicit parts.
There are examples of focussed research into the computational support for collaboration which can be examined --- such as that by Baker on negotiation [Baker, 1994], Dillenbourg and automated 'co-learners' [Dillenbourg & Self, 1992], Blandford and mixed initiative dialogue [Blandford, 1994], Soller and the use of sentence openers [Soller et al, in press].
It is hoped that computer support for collaboration can be improved if there is increased awareness of the model of collaboration that is being supported. This improvement depends on a corresponding improvement in our understanding of the different forms of collaboration that are possible, and their costs and benefits. Finally, it is important to be able to evaluate the effectiveness of a specific collaboration. This will be hard to do if there isn't a clear view of what kind of collaboration is being considered.
My thanks to my colleagues Mark Burton, Daniela Romano, Vania Dimitrova and Susan Bull. Thanks also to Patricia Tedesco for her comments on a draft of this paper.