Agent’s Contribution for an Asynchronous Virtual Classroom

Kenji Matsuura, Hiroaki Ogata, and Yoneo Yano Faculty of Engineering, Tokushima University 2-1, Minamijosanjima, Tokushima 770-8506, Japan http://www-yano.is.tokushima-u.ac.jp/ (matsuura, ogata, yano)@is.tokushima-u.ac.jp

Abstract. This paper describes a new idea about the multi-agents that

act as classmates of learners in an asynchronous virtual classroom. The environment of this virtual classroom is designed to support asynchronous participants. Some agents in the system are based on learnermodel of the past learners, and others are on the educational strategy the teacher has given in advance. In such situation, it is necessary for a learner to filter classmates who are actually software agents in the user interface. Furthermore, the actions of these agents are also to be defined. Therefore, we propose two types of strategies to select well-fitted classmates first things in this paper. Then, we define these agents’ action in an agent-based asynchronous virtual classroom.

1

Introduction

It becomes more common for those who design a lifelong learning environment to realize it on computers and networks as the virtual school/classroom [2][6]. These movements make the ordinary learning environment, like classroom-based education in a school, extend to the virtual one via the Internet [3][4][5]. However those environments could have some problems that the system has to compel users to participate in at the same time [8]. Otherwise, if the system allows a user to participate in anytime, all the things she/he can do in that environment is to see static learning materials for the reason why she/he has no way to communicate with others synchronously. The problem is that it is awfully difficult for an asynchronous learner to communicate with others immediately when she/he wants to ask or to make comments on the learning materials in the classroom. For example, concerned with a supporting distance-lecture system, learners have to use the same learning materials at the same time, otherwise the system cannot provide the collaboration space among learners. On the other hand, the approaches to bind software agents to a learner’s environment as a companion have been seen recently [1].

G. Gauthier, C. Frasson, K. VanLehn (Eds.): ITS 2000, LNCS 1839, pp. 344-353, 2000. © Springer-Verlag Berlin Heidelberg 2000

Due to the teacher's educational linking strategy. In the concrete. they have to participate in synchronously. 2 Agent-Based Asynchronous Virtual Classroom 2. AAVC system combines self-instruction with collaboration by agents. A learner in this system is able to see some classmates and their actions in the environment. if the classroom is opened to the public. Then the system sends its learning materials and additional information about the classroom on the educational strategy. In such situation. we also define their action in our AAVC system. or they may use asynchronous communication tools without any utilization of past learners’ action. we describe how to select these autonomous agents as classmates for a learner first of all. while the other real learners don't participate in the same class at the same time. In this paper.1 Features of AAVC It is important for those who learn in an asynchronous learning environment that the system provides a virtual community space for teaching and learning. Hence this system links people as well as machines across the time and place. if the system provides a collaborative environment to learners. A learner in AAVC system uses networked computers to learn or communicate with remote learning resources. AAVC has the efficiently visualized learning environment to support distance lecture that uses exposed video images of a teacher and some virtual learners. The learning environments of some ordinary studies propose one-way communication in which the systems show a learner teaching materials. Therefore. a learner can make his own knowledge amplify. The environment of AAVC has some comparable advantages shown as follows: (1) Anytime/Anywhere learning The system enables a learner to request the resources at anytime on WWW. (3) Learning by observing others' interaction Watching the interaction among other learners or between a learner and the teacher makes to construct effective discussing way for the learner and to be aware of . Some of them are software agents of past lectured learners and the others are on behalf of prepared educational agents. but without the requirement to be online at the same time. Then. On the other hand. including teachers and other learners. (2) Seamless Learning Other classrooms linked by the teacher are seen in an information space. we have been developing AAVC (Agent-based Asynchronous Virtual Classroom) to solve those problems. A learner in this system may also collaborate with other learners or a teacher by using agent communication. The information are may be changed by means of the strategy every time a learner participates in AAVC.Agent’s Contribution for an Asynchronous Virtual Classroom 345 The main purpose of the researches to support distance learning is to study how the system provides a collaborative learning environment for learners or makes a learner acquire new knowledge and refine it on network [7]. and a learner may moves to other classroom at anytime. a learner feels much presence that she/he shares time and space with a teacher and other learners.

(4) Learning through collaboration This system provides to learners a collaboration space with others. At the beginning of a classroom. who are divided two groups: past learners' agents and educational agents. the system indicates the existence of learner D. . When the learner D participates in the middle of the classroom. As the learning time goes on. The system indicates other learners’ past action. As we focus on certain learner. the AAVC might be opened to the public on WWW.2 Time flow Fig. without any interrupting the learner’s lecture. she/he can also recognize the other learner D. especially their statements in the past discussion. When the other teaching materials are ready for the class. the system organizes the classmates. she/he can see some agents of the other learners who have participated in at the past time. Then the system enables learners to request to participate in. actually agents. which is made into MPEG or simply RA file. At first.346 Kenji Matsuura.1 shows the time flow in AAVC. Time Flow 2. The system prepares synchronous / asynchronous collaborative environments for the continuous discussion. 1. and some additional agents in the user interface. a learner can refine his/her acquired knowledge through the discussion. In such situation. Hiroaki Ogata and Yoneo Yano various viewpoints. the lecture is recorded as the video image. (a) Synchronous Mode Teacher Learner A Learner B Learner C Learner D Focused Learner (b) Asynchronous Mode Absolute Time Flow Reflection NOW Video Time Refer Discussion Synchronous Discussion (with Learner D) Asynchronous Discussion (with Teacher. Learner A) Relative Time Flow Fig.

whose conceptual model is illustrated with Fig. Agents’ conceptual model The kinds of software agents are shown in Table. This agent plays a role that .g.1 Types of Agents Table 1. 2.Managing the learner’s individual knowledge ..Imitating other learner’s past action .Agent’s Contribution for an Asynchronous Virtual Classroom 347 3 Agents Model 3.Carrying out an action plan of the claque learner that the teacher have given in advance .Managing the other learner’s individual knowledge . e.2. clicking a mouse or typing a keyboard. The types of agents in AAVC Types My-agent Simu-agent Whose Agent? Learner-self Other learner Claque-agent Helper-agent Teacher (as learner) Teacher Each roles or explanations .1.Managing shared knowledge . My-agent displays some action patterns according to the learner's operation.Answering the learner’s questions Teacher (4)Helper -Agent Organization /Answer /Comment Social Knowledge AAVC System User Interface Learning Materials (3)Claque-Agent Question /Examination Action Negotiation/ Communication (2)Simu-Agent B’s Knowledge Question (1)My-Agent A’s Knowledge Learner A (Current) Learner B (Past) Fig.

The statement belongs to some kinds of question. each statement has to go to the tune to the relative time in each lecture. of which agent simulates the other past learner’s action to be aware of other’s view points and shoots some questions to the learner to improve the acquired knowledge into stable one. As to other statements. Hiroaki Ogata and Yoneo Yano monitors the learner’s action and manages his/her individual knowledge. Since AAVC system abbreviates the past discussion. but without stopping its discussion and the statement is added at the end of the discussion at the relative time.2 Agent’s Action In order to realize these agents’ action. implemented on the server-side scripts. of which action is generated on reaction-planning architecture. she/he must start new discussion with a new title. claque addresses to the learner and brings his/her answer to helper on the server-side. The first statement in a discussion goes at the relative time (RT) from the beginning of the lecture. Concerned with agents action on the client side. two functions are necessary. Simu-agent imitates the past learner’s actions and manages the learner’s individual knowledge. or a fixed form of an examination. . Since these actions make a learner have his/her own opinion or want to ask some questions to others about shown statements.348 Kenji Matsuura. comment. what statement to appear. So the system enables a teacher to establish the claque-agent’s addressing with some attributes (relative time in the lecture. Each agent has its own action model. (2) Claque-agent’s addressing It is important to fix the acquired knowledge in the lecture to the stable. If she/he wants to cut into the discussion. (1) Simu-agents’ simulation This example of script shows how simu-agents imitate the past learner’s action when a learner requests to show certain past discussion. both actions may also induce collaboration with others. Claque -agent is a preset-typed educational agent that acts for the learner to give him/her some advice. When a learner attends at a lecture. there are some cases that a learner wants to ask some questions or to comment on his/her opinion during watching a discussion. 3. Of course. we have defined what kinds of actions are suitable in such environment at some points of view. In that environment. In those types of software agents. the system calculates the start time in the way to summate RT plus each statement’s calculated interval time. and other flags). Simu-agent acts on the past-attended learner's model. she/he can. the number of claque-agents is adjusted by the system. which is constituted by the learner's action log. Helper-agent acts on the pedagogical strategy given by the teacher and answer the learner’s question.

(1) Regard the communication among learners as important The system selects simu-agents as classmates at first. If topics are not found. who have participated in the classroom and uttered in this model. from the shared knowledge base. When the members in the classroom are few at the beginning.3 Classroom formation AAVC has two ways of classroom formation that are available in which a learner requests to join the classroom. the system attempt to support an asynchronous encounter and to increase collaborations among learners. In that case. While this type makes a learner collaborate with teacher rather than other . First role is that helper diagnoses the learner’s answer of the claque’s question and replies it if a teacher establish the action.Agent’s Contribution for an Asynchronous Virtual Classroom 349 While (exists statements) { If (first statement of the discussion) { Then { Define the agents’ action as follows: Start time: relative time (RT) of the Discussion Interval time: calculate time fit for the statement length Display: parallel (statement 0) } } Else if (statement ID == N) { Then { Define the agents’ action as follows: Start time: relative time of the statement Interval time: calculate time :(RT + summate later statements time) Display: parallel (statement N) } } } (3) Helper-agent’s advise Helper-agent plays two important roles. The second role is to retrieve candidates to the question of a learner. helper-agent retrieves from personal knowledge base. But the number of participants who have joined to a classroom increases as time goes on. (2) Regard the educational strategy of the teacher as important The system selects claque-agents as classmates to be given the high priority of selecting. the number of the displayed agents is large. In this way. The system gives high priority to be selected to those who have uttered more times than others. only the candidates from personal knowledge base of which attribute is open to public are shown to the learner. 3. Then the displayed classmates have to be filtered off on certain criterion as follows. which is inputted during watching the past discussion.

4 Implementation 4.1 System Configuration Fig. Hiroaki Ogata and Yoneo Yano learners. clients have only viewer and don't have any data. At the beginning of the classroom. the system use (1) because many topics among learners are expected to discuss then.350 Kenji Matsuura. 3.3 shows the system configuration of AAVC. the server is to control and the clients to request. AAVC changes how to select these criterions automatically. Each data and Modules on the server side are shown as follows: Shared Knowledge Personal Knowledge Action Log Agent Planning Manager Communication Server Contents & Templates Server (Servlet) Session/Time Manager Communication Middleware Client (Browser   & Applet) Time Keeper Communication Client Agent  ctor Action  onitor M Agent Actor  Agent A ctor A UI Learner Fig. System Configuration . Since the number of discussion occurred in the classroom decreases. As a fundamental rule. a learner may want to collaborate with other learners in order to answer the question from claques. AAVC system is built on the WWW with Java. While all available data/files are located on the server side. the system selects (2) to intervene the learner’s study.

A learner can use memo tool at anytime she/he wants. which will be processed and made into a learner model. Action Monitor: is to monitor the clients interaction and to propose additional information to a learner 4. and e-mail based asynchronous communication among learners. monitoring the synchronous communication with chat. other session variables and relative/absolute time of clients’ lecture Agent Planning Manager: is to control the agent's activity as keep the time Agent Actor: is to act each action thread in time order unless agent-planning manager gives priority to another thread. and then deleted when the learning finished. the AAVC system has the personal knowledge base to storage individual memo. 4. Action log: Action history log files of hold session. which is used for another learner. which shows a teacher's lecture with its teaching material. For example. Shared Knowledge: Shared knowledge base is constructed of some tables. The AAVC system makes into learner model by diagnosing these action logs. Session/Time Manager: is to manage the session-ID. . which can be edited by the teacher. One is video image frame. Personal Knowledge: Compared with shared knowledge. of which enables to write the memo text and some drawings.Agent’s Contribution for an Asynchronous Virtual Classroom 351 Video contents: Video images of the lectures. They are processed into individual objects in the database on server-side. question and answer about prearranged keywords. An example of User Interface There are two factors of technology to give learners the feeling of much presence.2 User interface Fig.

L. are brought to a learner by clicking the virtual objects and selecting displayed menus. Those menus make learners collaborate with others.. To realize that. (1998) Gerhard F. S. L. the system sends a question-message to learners or a teacher by asynchronous tools like e-mail. IBM Systems Journal. So our future work is to reduce the work of a teacher and to evaluate the prototype system through its trial use. which shows the other learners' action. In some cases. who are classified to a teacher. the teacher may confine the member of the class. a learner inputs some questions or comments to the WWW server. this paper specially emphasized on how the system selects multi-agents to construct the classmates.38.A. In addition. If any candidates of answer-sentences are not retrieved. Vol. L. However. NSF Symposium Proc. as she/he wants.: lifelong Learning. the teacher has to work hard for initial readiness.352 Kenji Matsuura.: LANCA : a distance Learning Architecture based on Networked Cognitive Agents. new technologies are making possible innovative approaches to educating students and employees outside of the academic and corporate classroom..: learning Networks. M. she/he goes into the virtual classroom of which mainly components are the lecture video image and learners pictures. Due to AAVC. extra information is seen in another windows to show the classroom's properties.4 (1999) 508-530 Harasim. The video image is embedded at opposite side to the virtual learners in the window. when those agents acts. R. real learners' information to participate in. but it does not in other cases.. Some certain actions. A learner can see the lecture in it and control the streaming as the learner's occasion demands. another learner. That's depends on the teacher's readiness for the classroom.S. No. the AAVC system plays very important role. Teles. E. the server rapidly replies at the actions of the learner. Hiroaki Ogata and Yoneo Yano and the other is classmates frame. and what kind of actions of them is activated. G. While we have not referred to the membership in the same class. C.. on Learning and Intelligent Systems (1997) 7-12 Gregory D. and the other classrooms' information. To collaborate with others. teachers and learners are needless to share time and place. Its architecture allows learners to learn anytime and anywhere. Gouarderes. and the system's educational agent. As the knowledge demands imposed upon those who teach and learn increase. Aimeur. The MIT Press .4 is an image of main window. Hiltz. and Turoff. of ITS'98. Proc. For learners who are geographically remote from tutors and peer students. Martin. which are carried on table 1. When a learner requests a lecture to watch.: Classroom 2000: An experiment with the instrumentation of a living educational environment''... Fig. 5 Conclusion We proposed the agents contribution in an agent-based asynchronous virtual classroom. U. Abowd. References 1 2 3 4 Frasson.

: Asynchronous Learning Networks as a Virtual Classroom. (1996) 232-237 Matsuura. Boston. No. and Davies. H.: LIFELONG LEARNING. of Educational-Telecommunications 1996.S. Proc.: Agent-based Asynchronous Virtual Classroom. H.A. K. Proc. Japan (1999) 133-140 . N.. S. and Wellman.40. U. K.1. Communication of ACM. K. Ogata.. and Yano.. of ICCE’99. and Yano.: Knowledge Awareness: Bridging between Shared Knowledge and Collaboration in Sharlok.. Y.. Kogan Page Ogata. Y. Matsuura.Agent’s Contribution for an Asynchronous Virtual Classroom 5 6 7 353 8 Hiltz.. Vol. B. W.9 (1997) 44-49 Longworth. Vol.

Sign up to vote on this title
UsefulNot useful