creating an organisational learning memory for

Teleteaching'98 Distance Learning, Training and Education (Proc. of the XV IFIP World Computer Congress, 31 August-4 September 1998. Pages 1035-1046

CREATING AN ORGANISATIONAL LEARNING MEMORY FOR COLLABORATIVE EXPERIENCES IN DISTANCE EDUCATION Verdejo, M.F. & Barros, B. Departamento de Ingenieria Electrica, Electronica y Control, Escuela Técnica Superior de Ingenieros Industriales (U.N.E.D) Ciudad Universitaria s/n, 28040 Madrid, Spain {felisa, bbarros}@ieec.uned.es phone: 34 - 1 - 398 64 84; fax: 34- 1 - 398 60 28

Abstract We are presenting an evolving system, which will allow distance learners to carry out activities in a collaborative way. We introduce the concept of organisational collaborative learning memory as a structured repository of a collection of group learning experiences. The goal is to use past learning activities as effective resources for further work. The architecture of the system is organised in three levels, a generic component, an application level, and an organisational one. The system provides both the means to capture learning outcomes and processes, and a tool to organise them for further reuse. After describing the architecture, we discuss a number of directions that have been experimented on in different collaborative applications along the design of the system.

1. Introduction and background Collaborative learning scenarios engage learners in a variety of cognitive and metacognitive activities to promote the conscious cooperative development of shared knowledge. The technology base for these kinds of environments includes a diversity of hypermedia database tools to support remote social knowledge construction for a variety of task models. CSILE [11] and Notebook [8] are two examples of school oriented collaborative systems, focusing on theory building and scientific inquiry respectively, while JITOL [6] targeted a more professionally oriented population. This latter one included a combination of conferencing tools and annotating tools. The first one, to support the interaction and sharing of information between experts when solving a case together, and the latter in order to help them to reify their professional expertise to build a communal database for further practitioners' re-use. We can consider a knowledge pool as a shareable description, capturing the common understanding reached by the learners in their discourse. This knowledge is to be dynamically


enriched with new contributions and connections from learners. The tools assist learners in representing and structuring their own knowledge and supply ways to create and incrementally add new contributions to the communal database, linking them explicitly to other contributions. A further extension is to conceive the concept of organisational collaborative learning memory i.e. a structured repository of a collection of group learning experiences. The final goal is to provide access to the recorded events so that the activities could be revised and used as effective resources for further work. We foresee at least two main uses of this kind of memory (1) to offer students additional resources for their learning tasks and (2) to facilitate the task of defining new learning experiences partially based on previous ones. We present an evolving system to support collaborative distance learning. Evolving in the sense that learning events are a source to create and enrich an “organisational learning memory”, to be exploited as mentioned above. For that purpose the system provides both the means to record learning outcomes and processes, and a kind of authoring tool so that learning events could be collected, selected, contextualized, and organised for further re-use. The structure of the organisational learning memory and the authoring tool are described in section three of this paper. But first of all we are going to discuss the overall aims and the general system architecture.

2. Aims and Overall System Architecture Our aim is to explore collaboration at a distance for a range of tasks involving complex interactions. We focus on distance learning scenarios where students are geographically distributed [3]. They work collaboratively in small groups to carry out learning activities previously designed by a teacher. Asynchronous communication is used; there are two reasons for this choice: distance learning students rarely have compatible schedules to permit real time sessions, and the technology required is still too expensive for our target users. Furthermore, the kind of learning tasks we are considering need thought and reflection, and hence asynchronous communication is well suited because it allows each student to work at his own pace. The use of Computer Supported Cooperative Work (CSCW) technologies in collaborative learning opened a new application area, known as Computer Supported Cooperative Learning (CSCL) [10]. Many ideas where CSCW tools and methods could enhance the support for group learning processes have been pointed out [7][12]. We want to center our concerns on the following three key aspects: (1) mediation on human-human interaction, (2) task-oriented support for learners and (3) the creation of a collective memory from the activity of a community of both learners and teachers.


A CSCL activity can be recorded, observed and studied from two viewpoints: as a process and as a product. Both aspects are sources of information relevant not only for the participants but also for other learners and teachers. The idea of building a collective memory arose in professional environments to address the problem of providing collaborative help mechanisms. Answer-Garden [1][2], and the version running in Web, named Answer-Web, are well-known systems in the field. The problem is to define workable approaches to enable the growing of this kind of collective learning memory i.e. to design procedures to support a distributed authoring process to build a knowledge pool, where relevant information could be made easily retrievable for further needs. Our approach to tackle this problem is to exploit the structure of the learning events. We will elaborate on this idea once the system architecture has been described. 2.1 A Generic System for developing Cooperative Distance Learning Experiences In designing software to support group learning it is crucial to have systems that are open to adapt to changing demands and can grow incrementally to handle new situations. An explicit model at an adequate level of abstraction allows for changes without substantial programming effort. We have characterised classes of collaborative generic learning tasks in terms of three interrelated components: a group of learners, a task and a collaboration procedure, each one respectively modelling the relevant features for designing the structure and support to carry out computer-mediated collaborative activities. A range of new situations can be modelled by reusing

TASK Outcome Description Learning Resources Tools for working Subtask Structure

GROUP ROLES

COLLABORATION PROCESS Communication Units and Structure

PREDEFINED WORKSPACES

GROUP MEMBERS

SPECIFICATION

LEVEL

WORKSPACE

EXPERIENCE

LEVEL

ENVIRONMENT TO SUPPORT A COOPERATIVE LEARNING EXPERIENCE ORGANISATION

LEVEL

ORGANIZATIONAL LEARNING MEMORY only 1 1 or more

Figure 1. Three levels of the system


and combining the above defined generic learning tasks in order to generate collaborative workspaces. Our system is organized on three levels (see figure 1): (1) The Specification level. It includes the generic components and methods allowing to automatically generate either an individual or a collaborative workspace for a particular learning task. An authoring tool is provided to interactively define the requirements to configure a workspace for each particular activity. The system automatically generates a customised workspace to carry out the collaborative process. Later on, an application can be built by combining several workspaces, so that all the functionalities appear integrated into a personalised interface for each learner. (2) The Experience level, where a tailored system will be available to support a group of distance learners to carry out a collaborative learning activity. (3) The Organisational learning memory, where learning events are captured and represented as a structured collection of cases. The content can be reused for learning purposes in different ways. 2.1.1 Task and conversation models The workspace structure depends on the kind of collaborative learning task. Collaboration is conversation-based. Conversation consists of turn taking, where each contribution both specifies some content and grounds it [5]. The type of contributions and their constraints can be defined to establish an explicit structure for a conversation. For example, argumentative coedition consists of an interplay of proposals/counterproposals between peers, refining ideas until reaching an agreement. We have characterised a set of tasks by which we can define a range of collaborative applications such as “synthesise” (which consists of reading the working material and making a summary), “compare and value” (a task consisting of analysing information sources to compare and evaluate them), “search and organise” (a task for retrieving and organising material about a given topic), and “complete” (a task consisting of completing the ideas of a given text using working materials). Each generic task is described in terms of a set of features. For instance a collaborative synthesis is fully specified by (1) the available multimedia material (papers and other sources to consult), (2) the working resources such as word processors, file managers, or search engines (3) the type of outcome to be created, for example a document structured by topics, (4) the substasks involved,


their temporal order and their dependencies on the outcome, and (5) the type of dialogue to achieve the subtasks collaboratively, for example argumentative coedition finished with explicit agreement. Our underlying model for a conversational structure is a labelled oriented graph. This mechanism allows structuring a group conversation in a generic way. The nodes represent conversational unit types and the edges represent the allowed links between the different types of contributions while a conversation is going on. For our framework, a categorisation of types and conversational moves are enough to support the communication requirements. The content is to be interpreted by the participants. A semi-formal approach allows us to achieve a generic intermediate level, both useful and flexible to accommodate as much or as little structure as needed for a particular kind of application. For learners the benefit is twofold (1) they receive some support for the process of categorising and organising their ideas when contributing to the debate and (2) further inspection of the process is facilitated because the system can take into account the type of units. For instance visualisation operations with filters can be easily added to observe the process that the group follows for reaching a goal 2.1.2 Configuring a Group A group is configured in two steps. At the specification level, the kinds of roles are defined. Group members and the mapping between members and roles for a particular experience are defined in a separate mode [9]. This way of configuring the groups allows us to reuse the workspaces for a number of similar learning experiences but with different participants. A role captures the rights and responsibilities of the participants in the learning task, for example the active contributor, moderator, questioner, monitor, etc. A role encapsulates the set of permissions to operate in the learning environment such as accessing, reading, contributing, or resuming the workspace. Further details about the way to define roles and groups in our system are given in [4]. 2.2 Developing an application To create a collaborative learning environment we can combine instances of working spaces when it is required for a learning task. Next we will describe one of our case-studies. This environment has been used to run a Distance Education PhD AI Course in the 1996/97 academic year. The environment could be accessed through our department Web server, and students were using the system via Internet from different places in Spain.


The collaborative learning task was reading and understanding a collection of papers and then elaborating a synthesis of the main ideas through a common discussion process. Students were recommended to go through two main phases; a first phase centered on an individual reading and understanding of the material; a second phase consisted in elaborating an essay in collaboration. For the first phase, personal work, communication with peers was also encouraged. Doubts, clarifications and further information could be interchanged within each student group to facilitate them a better understanding of the material. In the second phase students had to rely on their understanding of the subject to elaborate a synthesis of the papers. An important issue in this scenario was the planning and scheduling of the group assignment. Students had to agree on a schedule and be able to monitor their progress so that they could finish their work in time to fulfil the academic deadlines. Thus the kind of learning task was argumentative coedition; however some coordination of task management was also required. For this application the following shared workspaces were created: a “coedition workspace”, where students could carry out the joint writing of their essay; a “coordination workspace", where task management issues could be discussed and agreed on, and two predefined workspaces: a “version workspace” for storing coedition space releases, and a “result workspace” where the outcome for each section was stored automatically after an explicit group agreement. The coedition workspace was defined by using the system specification authoring tool. Data acquisition processes were carried out to include the following specifications: •

Group structure: two or three students and a tutor. Students play the role of contributor and the tutor plays a questioner role.

•

Task •

Multimedia material: a collection of papers and a collection of URLs.

•

Working resources: a word processor, import/export file facilities, access to the Internet

•

Outcome structure: an essay (a decomposable document), headings of the essay sections defined by the tutor.

•

Subtask structure: mirrors the document structure, as many subtasks as sections, no ordering required.

•

Collaboration Process •

Conversation structure: a labelled oriented graph describing the conversation flow.


Link to Individual Space Link to Version Space Link to Coordination Space Contribution name and creation data

Task structure

Link to Result Space

Contribution text

subtask

Contribution data and content Links to contributions related with PROPOSAL

Contribution link

Contributions related with proposal in the conversation structure

Form for replying to this contribution

Text area for writing the new contribution content

Figure 2. Coedition Workspace Interface •

An agreement method: explicit consensus about a proposal or contraproposal for each section of the outcome.

Figure 2 shows the learner interface for the coedition space generated by the system from the above definitions. Task structure is represented on the left area. At the beginning, there is only the list of subtask titles. As the process goes on, dynamically, for each achieved or active subtask the conversation process is represented by the indented contributions appearing below the subtask name. The system does it automatically whenever a new contribution occurs. When clicking on a contribution the name of the subtask, the full contributions, and links to related contributions, will appear in the upper right part. The bottom half of the right area allows us to create new contributions. Selecting the type of contribution from a list of choices, where just the allowed categories are displayed, the text can be directly edited or imported from a text or HTML file.


Once students reach an agreement on a proposal for a subtask the system automatically writes the content into the section of the document related to that subtask. Clicking on the "Result space" checkbox, students can visualise the current content of the document. The system supports document versioning and provides a voting mechanism for achieving an agreement. The coordination space is a shared area where the group members can send public messages for all members of the group, or private messages either for the teacher or for other group members. A private agenda and a group agenda are also provided. Further description of these features is given in [3].

3. Creating an Organisational Learning Memory from experiences The content of workspaces can be seen as a source of new material for other learning experiences. All this information can be organised and stored in a case-based library that we call organisational learning memory (OLM). It is a dynamic repository in the sense that its contents will evolve with the cooperative learning system use. If provided with efficient methods for augmenting, maintaining and accessing the information included in the OLM, both teachers and students could benefit from its use. For teachers it is interesting for generating new examples for analysing the groups’ way of working, for creating new material, etc. For students it would be useful as direct study material, as working material for other cooperative learning experiences or as examples of ways of working (or as ways of not working) cooperatively. Workspaces capture the collaborative process and the outcomes of learning experiences. In order to reuse this information some elaboration is needed to collect, select, organise and assess what should be included as a case in the OLM. Our strategy has been to support the current work practices and then to plan a series of improvements of the tool along with further use of the system. By now, cases have been elaborated by designers and instructors. Cases can be defined at different granularity levels. We have found it useful to establish three levels (collections, units and subunits) and two types of subunits (outcomes and processes).


Figure 3. OLM Interface for adding a new subunit to the OLM

A case is built from a source (a learning experience) by selecting workspaces and annotating relevant issues. A case has a name, a source, and a set of features. Some of these features have to be defined while others are optional. For instance, Keywords, is a mandatory one, for each case a set of keywords should be given, indicating for each one a value (1 to 10 to the degree of relevance of that keyword for the case at hand; 1 being minimum and 10 being maximum). Keywords are useful for searching and connecting different cases as well as controlling the contents of the OLM [13]. Cases are created by interactively filling a set of forms provided by an authoring tool. An example is shown in figure 3. A subunit is the simplest case that could be stored in the OLM, and it keeps the contents of one workspace. Some subunits from the same experience form a unit. A collection means a group of related units or related subunits (unmixed) obtained from different experiences. A subunit is characterised by a set of features, some are automatically provided by the context (author, creation data), others are defined either by selecting options from a menu, or filling them


with appropriate values. Some features identify the subunit (name, experience source and workspace source) while others are for retrieval purposes: a list of keywords, or the workspace type (Outcome / Process). The rest of the features provide context and evaluation for reusing the case: the group information (number of members and background), task description (a description of the task the students have performed), assessment (a value in a qualitative scale to indicate how good the case is considered and the reasons to justify that consideration), and comments (an available free-text space where the author can annotate any information he considers interesting). The interface for adding a new subunit is shown in figure 3. A unit can have the same features plus a content feature (list of subunits usually the outcome and process subunits of one experience) A collection is characterised by the author and creation data, keywords and its contents as a list of units or subunits from different experiences. It has also has a grading field for evaluating each one with respect to the others, best case, worst case, ranking. The system offers a tool to help the author to create a case from the contents of workspaces related to a learning experience. Both units and collections can be added from the learning experience workspaces' contents to the OLM using this tool. Typical database operations are available such as searching information by a combination of features. Searching methods include keywords with a threshold value, related words, same task type, etc. The OLM can be used for different purposes. We have explore the following: •

The instructor can make an outcome unit explicitly accessible to students, for instance including it in the source material when defining another learning experience.

•

A collection of cases can be reused by the instructor for defining another kind of learning experience, for example collaborative peer evaluation of essays previously written by other students.

•

As material for further analysing learning experiences, to understand the way of working and cooperating within a group, and the variety of approaches. This reflection process can involve instructors and students. As mentioned above the system provides some facilities for browsing the material, for example the whole schema of the conversation or the number of proposals/contraproposals for a particular section of the outcome, can be visualised by different groups.

•

As material for evaluating the design of a learning experience, for example the adequacy of the task structure, roles and conversational units provided. That last potentiality is critical to


integrate users feedback, this is why we have included the set of features in the source field of a case shown on the small window of figure 3. The evaluation of each component used to define the learning task can be explicitly represented. We are exploring ways to use this information to improve and enrich the generic component of the system. •

As a source of advice for designers and instructors to define and develop other applications.

4. Concluding Remarks In summary, the OLM opens a variety of directions, providing new ways to support and facilitate the integration of past experience in the system for the benefit of future users. While we think this is a great potential to improve the whole process of creating and carrying out systems to support collaborative distance learning activities, we are also aware that many issues remain either problematic or controversial. Some of them are of a methodological nature such as constraining the learners' discourse, others of practical order such as the risk of overloading the task of instructors. Field trials have given us insights as to where the technology has been perceived as useful. Wider experiences will help us to focus and refine those ideas that could prove to be of interest for our academic community.

5. Acknowledgements The work presented here has been partially funded by CICYT, The Spanish Research Agency, project TEL97-0328-C02-01.

6. References [1] ACKERMAN, M.S., (1994), "Augmenting the organizational memory: a field study of Answer Garden" in Proc. CSCW’94, pp. 243-252. [2] ACKERMAN, M.S., MCDONALD, D.W., (1996), “Answer Garden 2: Merging Organizational Memory with Collaborative Help”, in Proc CSCW’96 (Ackerman, M. editor) November 16-20, Boston, pp. 97-105. [3] BARROS, RODRÍGUEZ-ARTACHO, M.,VERDEJO M.F. (1997),"Towards a model of Collaborative Support for Distance Learners to Perform Joint Tasks" in The virtual campus: Trends for Higher Education and Training, Proc. 3.3 & 3.6 IFIP 1997 Working Groups, Verdejo, M.F. & Davies, G (editors) Madrid, Spain, 27-29 November, pp. 123-136. [4] BARROS, B., VERDEJO, M.F. (1998), “Designing workspaces to support collaborative learning”, in IEA-98-AIE. (11th International Conference on Industrial & Engineering Applications of Artifical Intelligence & Expert Systems), Springer-Verlag, in press. [5] BOBROW, D. (1991), "Dimensions of Interaction: AAAI-90 Presidential Address" in AI Magazine, Vol. 12, No. 3, pp. 64-80. [6] BODER, A., (1992), “The process of knowledge reification in human-human interaction”, in Journal of Computer Assisted Learning, 8 (3), 77-58.


[7] COLLIS B. (1994) “Collaborative learning and CSCW: research perspectives for interworked educational environments”, in Lessons from learning. R.Lewis, P.Mendelsohsn (Eds). North-Holland. [8] EDELSON ,D.C., PEA R.D., GOMEZ L.M. “The collaboratory Notebook” in Communications of the ACM, April 96, vol 39, nº4.pp 32-33. [9] GREENBERG, S. (1991) “Personalizable groupware: Accommodating individual roles and group differences”, in The 2nd European conference on Computer Supported Cooperative Work (ECSCW’91), Bannon, L., Robinson, M. & Schmidt, K. (Eds.), Kluwer Academic Pub., pp.17-31 [10] McCONNEL, D., (1994), Implementing Computer Supported Cooperative Learning, Kogan Page Publishing. [11] SCARDAMALIA, M., BEREITER, C., (1994), “Computer Support for Knowledge-Building Communities” in The Journal of the Learning Sciences, Vol. 3, No.3, pp. 265-283. [12] VERDEJO, M.F. CERRI S. (Editors) (1994) Collaborative dialogue technologies in distance learning. Computer and System Sciences vol 133, Springer-Verlag. [13] WARD, D.R. (1991) “Boosting Connectivity in a Student Generated Collaborative Database”, in The 2nd European conference on Computer Supported Cooperative Work (ECSCW’91), Bannon, L., Robinson, M. & Schmidt, K. (Eds.) Kluwer Academic Pub., pp. 191-201.