Towards personalised learning communities on the Web Jesus G. Boticario*, Elena Gaudioso*, Carlos Catalina**
Abstract University distance learning is benefiting from the intensive use of Internet resources. This development is changing the teaching model. Lecturers and students are becoming active members of virtual educational communities. Thus communication and collaboration between lecturers and students is being enhanced. In this paper we describe a web site particularly appropriate for the development of cooperative work by the members of one group. Finally, we describe WebDL briefly, a multiagent system within this web site, aimed at personalising response to the user through machine learning techniques. Keywords On-line Learning Communities, Adaptive Web-based Systems, Distance Learning, Multiagent architectures, Machine Learning. Contact Information * {jgb,elena}@dia.uned.es Departamento de Inteligencia Artificial Facultad de Ciencias, Universidad Nacional de Educacion a Distancia, UNED C/Senda del Rey, 9,28040, Madrid, SPAIN phone:34-1-3987197/7242. Fax:34-1-3986697 **
[email protected] Unidad Tecnica de Investigacion y Formacion en Recursos Tecnologicos IUED - Edificio Facultad de Psicología UNED Ciudad Universitaria 28040 - Madrid, SPAIN Fax: + 34 91 398 66 93
1. Introduction It is obvious that positive use of the Internet media currently available will radically change teaching/learning relationships. The lecturer will have to become an information facilitator, a critical analyst of knowledge, a study guide, a reviser and assessor of a student’s academic education. Students will have to start to be aware of their essentially active role in the learning process as members of a virtual community of people with shared educational interests. Related to this aspect, but moving away now from the distance learning framework, there are more and more initiatives on the Web based on the development of virtual communities with different interests: communities of people with different levels of experience and with common interests, who come together on the Internet to take advantage of communication, collaboration and coordination with one another (Third Voice [2], Gooey [3], photo.net [4]). These communities use tools which facilitate on-line contact among users who are browsing the same pages (Gooey), foment the creation and management of different interest groups (Third Voice) and tested software (arsdigita [1] ) for the development of sites based on the storage of site information in databases affording, among other things, the following services: generation of dynamic page information, extension of their contents via user on-line annotation introduction, identification of user types and register of navigation traces. It is however becoming increasingly difficult to meet the wide variety of user demands and there are also the problems of inadequate structuring and the dispersion of Web sources (newsgroups, mailing lists, different kinds of pages such as those of the institutions, those containing courses, FAQs, the lecturers’ personal pages, practical exercises, distance learning assessment tests...). Naturally, this poses an obstacle to access to the information or desired service. In addition, the use of the resources already available depends a great deal on the degree of user training in their handling and knowledge of the web site structure.
Similarly, due to the increasing number of users who access the web sites supporting the creation of virtual communities, more and more problems are being detected in the management of these communities. Thus, mechanisms should be established to reduce the difficulties caused by different user interests and profiles. For example, help tools could be developed to connect people with the same interests or similar profiles. Bearing in mind these considerations, our objectives are very varied in nature, since they consider the quality of the teaching process and the effectiveness of on-line access. They are as follows: • To create a specific platform from and for Internet Distance Learning. • To foment significant and active learning by stimulating student participation in the use of the different resources. • To improve and ensure the procurement of the most relevant information and the establishment of the most significant communications for each user type by providing quick, efficient and personalised access. • To promote new ways of communication which facilitate the establishment of workgroups of students and lecturers with common interests by increasing the flow of information between all the protagonists participating in the process. • To stimulate the use of the technological resources available to users. • To personalise user access and services. In this paper we describe a web site, based on an adaptation of the aD- ArsDigita Community System, which is called aLF, especially intended to support workgroups on the Web. We then show how we extend this proposal with WebDL, a web-based educational system within aLF, intended to adapt sources of information and the Internet services to user needs. Our proposal, aLF+WebDL, is based on a new conception of web sites for managing virtual educational communities which can manage workgroups on the Web (with forums, electronic mail, shared workspaces, services for personal notification of events, group task management...) (see [5] for related web sites built with ArsDigita) and in a multiagent system, WebDL, which enables all the foregoing elements to be personalised to the student from his/her interaction with the web site and thanks to the application of automatic learning techniques.
2. General Description aLFi (developed by a development team from the Unidad Tecnica de Investigacion y Formacion en Recursos Tecnologicos (Technical Unit for Research and Training in Technological Resources) at the Instituto Universitario de Educacion a Distancia ii ), is the application that supports a web site designed and constructed in order to facilitate the development of virtual communities of users with the same interests and especially useful in distance teaching as it enables the students belonging to a particular course to form part of a workgroup with their fellow students and lecturers. It is a web site for Internet distance learning which is being used for workgroup management on the Web and includes: facilities for the management of on-line courses, shared workspaces, services for personal notification of events, public and private calendars, shared interest links, interactive conversation and research group management... and a lot more utilities presented efficiently with state-of-the-art service and architecture technology. The versatility of this site is that the information that the user receives can be personalised and adapted to each individual and/or group of individuals since the reply is dynamic and generated from all the information stored in the database. This web site has been developed using the ArsDigita Community System [1] ; a tool package for building scalable applications based on the web. It is based on a database and a set of scripts which enable the construction of pages which are dynamically personalised for the user from the data on this user held in the database. aLF is composed of a Web server (AOL server), a database (Oracle) and a set of TCL scripts which collect the user requests and construct the pages dynamically depending on the data which they have on the users in the database.
Some of the services that aLF offers are as follows: workgroups, news, bulletin boards, document management, chats, calendar, presentations, project management and tools to contact other users. • Workgroups: An area where there can be such services as: newsgroups, common calendar, group file storage with version control and grant management for the preparation of documents in cooperative work environments, group chats and forums, FAQ publication systems, massive e-mail management to group members, group activity project, task and notification of event management, etc. The administration of each workgroup is done by the person in charge who may not be the web site manager. The setting up of these kinds of workgroups is particularly useful for distance learning since it has meant that associated centres are managed in aLF (local UNED university centres dispersed all over Spain are linked), university department, different degree subjects (graduate and postgraduate subjects), courses (ongoing training for workers and teachers, open learning), study groups for research students and groups (see Figure 1) • Newsgroups: They allow access to newsgroups via HTML pages (see Figure 2), automatically index the messages by categories defined by the manager or establish a notification system so that when answers occur users are automatically informed. This last service combines the best of a forum, ease of access based on a web interface, with the best of electronic mail, the possibility of receiving information of interest on demand. The participants in the forums have the information like the date when the forum was founded, the daily messages, the most active participants or rather all the messages from one participant. Some of the forums of academic interest for our educational communities are: exercise forum (it includes information on the contents and organisation of the exercises, the distribution of the necessary material (trying to offer different options for its development and additional existing documentation on the Net)), First solutions forum (we propose using this service organised by topics, in order to make students aware of the prior conceptual structure of a topic that they are beginning to study. Students can consult the questions and responses given by other fellow students to the topic’s introductory exercises prepared by the teaching team and updated according to student indications. Those students interested, without consulting the rest of the material available, can leave their own solutions. These solutions can provide other students with an initial approach to the topic and will reveal the limited prior knowledge of the problem in question); self-assessment exercises forum (with this utility the students can gather sets of problems labelled by topics and by sections of the syllabus. Each exercise identifies the difficulty of the problem so that students can classify their selection. Each of these problems has the solutions given by the other students. A mechanism establishes that first students send a file with their solution to a problem before they gain access to the solutions to the problem).
Figure 1 Study group set up in aLF for a course
• Bulletin boards: With this service aLF users can create news and messages for the rest of the community members without the need for electronic mail. The date of publication of the news, its expiry date and even the group of people for whom it is intended can be controlled. This is because the news is for the workgroups. It is particularly useful since the rest of the workgroup members can comment on the news and group interaction is thereby enhanced. • Chats: Internet group work is useful particularly from an asynchronous point of view, i.e., when the different collaborators do not coincide at the same time when they do their work. However, on occasions the distance and feeling of isolation can be counterproductive. With the chat tool small group chats can be organised, thereby avoiding the problem of overcrowded chats which make the service unintelligible. The advantage of this service in aLF, as well as gaining access via the HTML pages, is that all the conversations are recorded so that the lecturer can select relevant information from his/her conversation with the students and thus avoid repetition. It is therefore advisable to transfer the key chat issues to the forum. • Calendar: There are private calendars for each of the users, group calendars and a public calendar for all the aLF communities. The most commonly used service is the fixing of appointments within the workgroup.
Figure 2 Newsgroup for IUED • Document management: The user can manage all his/her documents on-line, by having them in a central place accessible to all those users that he/she wants. The system has a permit concession service so that each document can be read, modified or managed by a specific user, by a series of specific users or by all a workgroup. These users can access from anywhere with a navigator to see their files and documents, can add new ones, copy them, etc., and their earlier versions are stored. Furthermore, we can know the history of the document and thus see who has worked on it and the date when the changes were made. The latter is particularly useful in distance learning since it enables development in workgroups suggested by the lecturer; with this version control it is possible to see which student has worked on which part, how they have worked together, etc. Figure 3 shows the TutorT course development group storage area structure for UNED telematic tutors.
• Presentations: They allow, by just completing forms, to prepare an instantaneous presentation for the Web which is quickly downloaded with no laborious tasks of sending files to the server; they have automatic indexes and all the links necessary for navigation support. They are really useful for teaching since they allow both lecturers and students to specify certain contents on a few screens (HTML pages). They also have a further advantage because you do not need to know the web format to edit the screen with a pleasant and navigable visual structure. • Tools to contact the other users: there are a set of different tools enabling users to publish their own personal pages, to see which users are connected at the same time (they promote the use of the chat tool), to see which users are recorded and what their participation is in the community.
• Project management: It allows certain project tasks (and their manager) to be created. For each task, task or project assignments can be created which become the responsibility of the task owner. These assignments can be different types (bugs, improvement, etc.), have different levels of severity (high, medium, low, etc.). Moreover, for every area assignment, the person in charge of this area can request more information about the assignment, reject it and check its state (open if it still has not been resolved, etc.). From the point of view of distance learning, the usefulness of a project management tool as presented here is obvious, we can create projects for certain student workgroups, and tasks for these projects, so that students work in their corresponding area as they receive assignments, either from their lecturer or their fellow students. The group members can check their work on a list where all the group member tasks are described, as well as progress reports on each one of them. Each project can have forums, project marks and a task and assignment monitoring system, thereby creating a user-friendly project knowledge base which can be consulted and accessed from the web.
Figure 3 TutorT course development group storage area
3. WebDL Despite the fact that aLF enables pages to be built up depending on the user who is accessing them, when the user enters the community, he/she may feel somewhat overwhelmed by the amount of services provided there. One drastic solution to this problem is to suppress many of the services provided, and so guide the student in accordance with the criteria which have been pre-established. Unlike this approach, our focus consists of analysing the interaction of each specific user in order to provide him with personalised services, adapted to his needs and preferences. In addition, a comparative analysis is carried out among the various users, so that the needs already detected for each type of system user serve to anticipate the requirements of new users. In order to access all the relevant data resulting from each interaction of the
user, aLF makes an exhaustive register of the group of users, and WebDL acts on this database and learns the detectable needs and requirements. WebDL [1] behaves like a Web-based adaptive Educational System [6] and is thus able to adapt the page contents and links to the user model in different ways (Adaptive Hypermedia, AH). It also corrects and assists students in their learning tasks (Intelligent Tutoring Systems, ITS). For this aspect, some of the distinguishing features of the WebDL system responsible for the adaptation in aLF should be highlighted. In addition to the AH and ITS techniques, there are also the so-called Learning Apprentice (LA) system faculties. Thus less effort is required for personalised support of user decision tasks by extending the initial knowledge base with machine learning techniques (for this we have drawn on our own experience in the development of these kinds of systems [7]. Furthermore, the approach that we have chosen is wider than the one generally used for adaptive hypermedia [4] because personalisation of the system, as well as guiding the user with link annotation, changing the order of the page elements, also personalises some of the contents and any other web site resource. Another distinguishing feature of this system is that the adaptation tasks not only refer to the user model elements (navigation type carried out, contact preferences, level of experience in the field, etc.) but they also identify, using collaborative learning techniques, services which may be of interest for a user community. These services are all the resources and material accessed by a group of users with a similar profile. By including WebDL in aLF, the dynamic pages offered to the student are constructed from the data held on the student and kept in the database and from the recommendations made by WebDL according to the data on the student and this student’s interaction with the system. This personalisation is done through user modelling acquisition. User modelling is carried out by means of a set of agents (see Figure 4) which combine the values learned with different biases corresponding to different machine learning methods (C5.0, Naive Bayes, Progol, Backpropagation, Tilde and Autoclass) [2]. To implement this combination, a special type of agent has been introduced, the advisor agent, which learns the competence range of each of the other agents. A task is only distributed to those agents that have been proved to be competent in it [3]. What differentiates our proposal from all those presented in the field of adaptive systems is that, besides the values of the attributes which make up the user model being inferred, the very architecture of the modelling subsystem adapts itself to each task. The tasks done with WebDL are described in [3]. As regards the adaptive collaboration support task [5] (the objective of the adaptive collaboration support task is to use the knowledge that the system has on the students (stored in the user models) so as to constitute workgroups and put students with similar interests and levels of experience in touch with one another, etc.) (see Figure 5) in WebDL we distinguish different kinds of collaboration between students and lecturers (see [8] for related work): Collaboration via public annotations on the course pages; Collaboration via asynchronous communications of the course components (news, e-mail,...); Collaboration via the setting up of workgroups based on student similarities. All these points are crucial on a web site like aLF, given all the communication and cooperation opportunities on offer. The placing of students in workgroups according to their interests and skill level (both for the subject and service use) is fundamental in a medium like the one presented here. With WebDL it is possible to create workgroups according to the information on the students in the database (personal and academic information and information on their web site interactions). 4. Conclusions and Future Work In this article we have seen the description of aLF, a web site particularly suited to the development of cooperative work by the members of a group. Since service personalisation that adapts to user needs and preferences is essential, we extend that proposal with WebDL, a system capable of adapting to student needs based on their interaction. In order to access all the relevant data arising from each user’s interaction, aLF draws up a complete register on the members of a group of users. WebDL acts on this database by learning the detectable needs and requirements. What differentiates our proposal from all those presented in the field of adaptive systems is that, besides the values of the attributes which make up the user model being inferred, the very architecture of the modelling subsystem adapts itself to each task, i.e., the advisor agent selects the learning agent which is the most competent for each task.
As far as the architecture is concerned, our future goals include the improvement of the characterisation of problems within the system, probably through clustering techniques, and the improvement of the advisor agent so that it might learn the set of relevant attributes for each modelling agent and for each learning problem. As regards the application of WebDL in aLF, the tasks implemented to date are those involved in adaptive navigation and adaptive collaboration support, and our most immediate objective is to implement the other tasks which enable us to give support to the student as he/she learns by intelligent analysis of student solutions and interactive problem-solving support. Compose the training examples from the database
Database
Trainer Agent
The advisor agent takes the solution tracks from the database to learn the competence range of the modeling agents Initializing user model agent
Advisor Agent
The training agents sends the training examples to the modeling agents
When the responses of the user modeling agents and the solution tracks the advisor agent decides which is the best agent for a certain problem User Modeling Agent Tilde
User Modeling Agent C5.0 User Modeling Agent Backpropagation
User Modeling Agent Progol
User Modeling Agent Autoclass
Figure 4 Architecture of the user modeling subsystem
5. Acknowledgements The authors would like to acknowledge the helpful comments of John Mullen and Anita Haney, arising in the course of his language revision of this article. We also thank the entire Artificial Intelligence Department for providing support for this project.
Figure 5 WebDL screen with a recommendation to the user
6. Web References 1. ArsDigita: www.arsdigita.com 2. Thirdvoice: www.thirdvoice.com [2000, November 6] 3. Gooey: www.gooey.com [2000, November 6] 4. Photo.net: www.photo.net [2000, November 6] 5. Related web sites built with ArsDigita http://www.arsdigita.com/pages/customers [2000, November 6] References [1] Jesus G. Boticario and Elena Gaudioso. A multiagent architecture for a web-based adaptive educaticon system. In Seth Rogers and Wayne Iba, editors, Adaptive User Interfaces, Papers from the 2000 AAAI Spring Symposium, TR SS-00-01, pages 24–27. AAAI Press, March 2000. [2] Jesus G. Boticario and Elena Gaudioso. A multiagent architecture to support distance learning personalization on the web. In Jim Etheredge and Bill Manaris, editors, Proceedings of the The 13Th FLAIRS Conference (in cooperation with AAAI), pages 124–127, Orlando, Florida, USA, May 2000. AAAI Press. [3] Jesus G. Boticario, Elena Gaudioso, and Felix Hernandez. Adaptive navigation support and adaptive collaboration support in webdl. In Proceedings of the International Conference on Adaptive Hypermedia and Adaptive Web-based Systems, number 1892 in Lecture Notes in Computer Science (LNCS), pages 51– 61, Trento, Italy, August 2000. Springer Verlag. [4] Paul De Bra and Licia Calvi. Aha! an open adaptive hypermedia architecture. The New Review of Hypermedia and Multimedia, 4:115–139, 1998. [5] Peter Brusilovsky. Methods and techniques of adaptive hypermedia. In User Modeling and UserAdapted Interaction, pages 87–129. Kluwer academic publishers, 1996. [6] Peter Brusilovsky. Adaptive educational systems on the world-wide-web: A review of available technologies. In Proceedings of Workshop WWW-Based Tutoring at Fourth International Conference on ITS (ITS’98), San Antonio, TX, August 1998. MIT Press. [7] L. Dent, J. G. Boticario, J. McDermott, T. M. Mitchell, and D. T. Zabowski. A personal learning apprentice. In Proceedings of the Tenth National Conference on Artificial Intelligence, pages 96–103, San Jose, CA, 1992. MIT Press. [8] J. Vassileva, J. Greer, G. McCalla, R. Deters, D. Zapata, C. Mudgal, and S. Grant. A multi-agent approach to the design of peer-help environments. In S. P. Lajoie and M. Vivet, editors, Proceedings of the International Conference on Artificial Intelligence in Education, volume 50 of Frontiers in Artificial Intelligence and Applications, pages 38–45, Le Mans, France, July 1999. IOS Press. Available on-line at: http://julita.usask.ca/Texte/AIED’99.zip. i
www.iued.uned.es/iued/tecinfor (IUED:www.iued.uned.es) An Institution linked to the Universidad Nacional de Educacion a Distancia (UNED: www.uned.es) ii
