You are on page 1of 5

Session F1B STUDENT MODELING AND SEMI-ATUTOMATIC DOMAIN ONTOLOGY CONSTRUCTION FOR SHIECC1

Sofiane Labidi and Nilo Srgio2

Abstract. - We are desiging and implementing, within the SHIECC Project, a computer supported collaborative learning system applied to courses in Engineering and Mathematics. We address in this paper the problems of students modeling. and knowledge domain construction within SHIECC. The proposed models have the benefits to treat the specificity of cooperative learning and to remedy the limitations of existing models. Keywords : Computer-Supported Collaborative Learning, Student Modeling and Assessment, Knowledge Acquisition, Domain Model Construction.

fundamental for the effectiveness of any ITS. In the context of cooperative learning doing this is little more different and complex [10]. In this paper we present first a brief description of the SHIECC environment and its conceptual model (sections 2 and 3). In sections 4, we define our model for student/teams modeling and assessment. In sections 5 and 6, we present the techniques developed in SHIECC to support the domain model construction. Finally we evoke some implementation aspects before concluding (sections 7 and 8).

2.

SHIECC FRAMEWORK

1.

INTRODUCTION

Cooperative learning has been around for a long time. However the use of computer to support such activity is fairly new. Computer Supported Collaborative Learning (CSCL) is a new emerging paradigm that extends classical Intelligent Tutoring Systems (ITS) by introducing the concept of cooperation [11]. The essential feature of cooperative learning is that the success of one student helps other students to be successful [4]. This is opposite to traditional classroom, in which the competition for rewards implies that students success may reduce the chance of anothers success. In cooperative learning, students work together to accomplish a shared or common goal. This goal is reached through interdependence among all group members rather than working alone. Each member is responsible for the outcome of the shared goal [6]. Cooperative learning is important because it produces greater student achievement than traditional learning methodologies. Beyond the academic benefits, we distinguish the social benefits (development of the student social skill), the economic benefits (less time and material are needed), etc. In this sense, we came up with the SHIECC project [7] [9] whose aim is to define a really cooperative intelligent tutoring environment. We are addressing here the problems of student modeling, student assessment, and domain model construction in the context of cooperative learning within SHIECC. All of theses concepts are depending and
1 2

The SHIECC framework presents the feature of integrating an intelligent tutoring system within a computer networking. Beyond the tutoring functionality, the SHIECC enables control of the interaction between the actors of a cooperative learning sessions i.e. the artificial tutor represented by the system, the students and the teacher.
SHI

Cooperative area 1
SHI SHIE

Cooperative area 2

SHIECC

Figure 1. Shiecc framework Students are divided in distributed groups acting as separated teams (cf. Figure 1). Each team consists of two or three interacting students with a terminal. The student team and their terminal within a physical space constitute what we are calling cooperative area. The teacher with its terminal constitutes a specific cooperative area. We define then a learning session as the interaction of several cooperative areas linked within a network and which could be distant physically. The network (an INTRANET) consists of a server and several PCs clients. These microcomputers are related using

SHIECC stands to Computer-Supported Cooperative Learning Hipermidia System in Engineering. Many thanks to CNPq for the financial support. Electrical Engineering Department, Federal University of Maranho, 65080-040 So Luis MA, Brazil, labidi@ufma.br

0-7803-6424-4/00/$10.00 2000 IEEE October 18 - 21, 2000 Kansas City, MO 30 th ASEE/IEEE Frontiers in Education Conference F1B-14

Session F1B
an Intranet. The students within a team work together at the same terminal. They collaborate with each other on the course topic using the computer as an active tutor and as a mediating tool for interacting with the teacher and other student teams in other cooperative areas. This interaction is done using communication windows (i.e. CHAT) enabled by the Intranet resources and the SHIECC system. So the students can ask for advice, make comments, send suggestions about the lesson, etc. The cooperative learning activities in SHIECC are composed of six phases (cf. Figure 2): (1) preparation of the student groups, (2) presentation of the knowledge, (3) course assimilation, (4) application of the acquired knowledge, (5) team evaluation, and (6) individual evaluation. Each phase has its specific functions and is contemplated with some appropriate pedagogical strategy to be selected [2] in agreement with the students team model [8].
COOPERATIVE PHASES PHASE 0 Group Preparation PHASE I Knowledge Presentation PHASE II Knowle dge Assimilation Assimilation PHASE III Knowledge Application PHASE IV Group Evaluation PHASE V Individual Evaluation

Figure 2. Cooperative pedagogical learning activities in SHIECC

3.

CONCEPTUAL MODEL OF SHIECC

The proposed cooperative intelligent tutoring system consists of the main concepts of tutor, pedagogical, domain model, and the compound student model (cf. Figure 3).

Pedago gical

TUTOR

Domain Teacher Interface student model update Student answer Evaluation


Compound Student Model

The tutor, interacting with the domain model, transmits the lesson to the teams of students. The system evaluates constantly the behavior of the students during the course progress (in different levels: knowledge presentation, knowledge application, etc.). The student team model is updated regarding to the result of the diagnosis/evaluation process. We note that in the case of individual evaluation only the student model is updated and not the team model. The tutor collaborates with the pedagogical model to decide the new cooperative learning strategy to be adopted afterward. This is depending on the behavior of the learners (which is represented within the student/team model). The SHIECC system is modeled as an agent-based society [9] [3].

4.
Students Interface Team model

STUDENT M ODELING

System answer

Figure 3. Conceptual model of the system We note the definition in SHIECC of two different kinds of interfaces: the student interface and the teacher interface. This is justified by the distinction between the functions and access rights of students and the teacher. The control between all these concepts could be considered as an important component of the system. The Figure 3 illustrates also the basic kinds of interaction within the system.

The Student Model (SM) has the aim to represent the dynamic behavior of the student while using an STI in order to optimize its decisions and actions. Most of existing SM are based on the overlay and bug models or the perturbation model. The overlay model focuses on the comparison between the student knowledge and the domain knowledge. The bug model is based on a library of possible mistakes that could be made up by the student in its pedagogical activities. The perturbation model however is an hybrid model that involves the concepts of the overlay and bug together. All of these models present some limitations in both conceptual and implementation levels. Moreover they dont address the paradigm of cooperative learning as adopted in SHIECC.

4.1 Compound Student Model The SM proposed in SHIECC is an hybrid one. It is based on the junction of the overlay and bug with the benefits to be applicable to cooperative learning and to minimize the 0-7803-6424-4/00/$10.00 2000 IEEE October 18 - 21, 2000 Kansas City, MO 30 th ASEE/IEEE Frontiers in Education Conference F1B-15

Session F1B
limitations of the existing models. This is made possible especially valorizing the teacher's role in this process. The model, denominated Compound Student Model (CSM), surveys the changes happened in the student individual and team behavior. It compares the current profile of the student/team with the initial profile to check the real improvements acquired along a learning session. Thus, the students' quarrels can be detected and processed properly some times by the system and if necessary by the teacher who is considered as a fundamental ally in the evaluation task. The results of the evaluation process are used by the pedagogical agent as decisive factors for decision making (apply new strategy? tutorial feedback? etc.). As shown in Figure 3, the CSM consists of the student individual model and the team model. Such division has the objective to manage separately the individual student performances and the performances of the groups they belong. In fact when a group is evaluated to have a certain level of performance, this dont imply that their members have the same performances level. The CSM is updated based on the evaluation module output which compares the answers proposed by the student/group with the answers proposed by the system (i.e. the domain model).
Perfil to Teacher
Cognitive Features Right Answer Wrong Answer

class content that the student/team studied and obtained satisfactory results. As will be shown in the next section, a class content makes the elementary division of the domain model in SHIECC. Each class content is associated to one or more cycle of activities to be performed by the students. The student cognitive characteristics inform the student initial and actual levels. This could be for example inexpert, novice and expert (cf. Figure 5).
Student/Team Level Cognitive characteristics Initial Level Inexpert Novice Expert Actual Level Bad {Inexpert } Regular {Novice} Good {Novice} Very Good {Expert}

Student

Figure 5. Evaluation component The initial level is obtained through a preliminary questionnaire and/or an interview applied in the first phase (group preparation phase) of the SHIECC pedagogical process (cf. Figure 2). The success of the system depends on the good interaction among these elements. 4.2.1 Individual evaluation Students take individual quizzes to be evaluated. This could be done in different ways. The paradigm of cooperative learning is important to accelerate and facilitate the knowledge elicitation process. However, the individual student is still the main objective of SHIECC. This phase provides an indicator to measure the degree of success of the system. 4.2.2 Team Evaluation Three concepts are essential for a realistic group evaluation [6]: team rewards, individual accountability, and equal opportunities for success. The teams arent usually in competition to earn rewards. Individual accountability means that team success depends on the individual learning of all members. The team members are evaluated individually but they are rewarded as a group. Equal opportunity for success means that students contribute to their teams by improving their own past performance. This feature ensures that the contribution of all team members will be valued.

COGNITIVE CHARACTERISTICS RIGHT DATABASE MISTAKE DATABASE HIPOTHESIS DATABASE

SPECIFIC KNOWLEDGE

KNOWLEDGE TO BE VALIDATE

Hypothesis

Figure 4. Student/Team Modeling 4.2 Evaluation Process

When in many cases, the system could not evaluate the student answer based on a comparison between the domain model and the answer, this won't be considered as a wrong answer but an hypothesis which will be forwarded to the teacher for evaluation (cf. Figure 4). Consequently the domain model and the two databases 5. DOMAIN M ODELING will be updated. This will allows in future situations, those same student answers don't need more to be evaluated. In SHIECC, we are investigating strongly the quality of For example, if a student is asked how many players the organization and representation of the tutoring make volley-ball team?. A correct answer is six players. knowledge (i.e. the domain model) and the problem of its However the answer two players should not be rejected acquisition from multiple specialists (teacher). This is since in beach-volley have two players. fundamental to attempt the effectiveness of the tutoring In addition to these components we distinguish the process. The domain model is fundamental to the student cognitive characteristics and the specific knowledge. As modeling module. output the evaluation component indicates the analyzed student/team profiles. The specific knowledge component stores the set of 0-7803-6424-4/00/$10.00 2000 IEEE October 18 - 21, 2000 Kansas City, MO 30 th ASEE/IEEE Frontiers in Education Conference F1B-16

Session F1B
For example, some concepts in an algebra lesson are circle, rectangle, triangle, etc. These have as attributes elements like center, root, height, width, etc. As methods and expressions on these concepts we have to define the area,. perimeter, etc.
Knowledge Contents Cycle of activities Generic Queries KB

Specific Domain Curriculum Units

Figure 6. Domain model The domain model (cf. Figure 6) consists of a set of interlinked pedagogical units that constitute the curriculum. The relationship between the units defines the pedagogical order of the tutoring process. Each unit of the curriculum consists of the knowledge contents and a set of tasks and problems to be solved by the students named cycle of activities. The knowledge contents constitute the pedagogical knowledge to be presented in the third phase (knowledge presentation, cf. Figure 1). Because of the student profiles (given by the student model) are heterogeneous and that different pedagogical strategies are possible in knowledge communication certain knowledge contents could/should be taught in different levels depending on the categorization of the students. Thats why in SHIECC the domain model has another dimension: the depth. This enables the pedagogical knowledge to be stored and presented in different levels and strategies.
Inquiring/ Structuring

Teacher

Scheduler

KC Plans

KC Design

Student

Application/ Evaluation

Knowledge Contents

6.

DOMAIN M ODEL CONSTRUCTION

The domain model is one of the main components of an ITS. However it constitutes the most challenging task in the ITS construction. This is because generally most ITSs dont offer facilitators to assist this process and because teachers, coming from different areas, do not have enough technical knowledge to perform this task. In this section we will give a succinct description on how this could be done in SHIECC. Our approach is based on the definition of a new module enabling a semi-automatic knowledge acquisition and domain model construction. This process is illustrated in Figure 7. The domain model is defined and structured in a specific Knowledge Base (KB) with the participation of specialists in the area (the teachers). The knowledge acquisition module interacts with the users (teachers) applying a generic questionnaire/inquiries for detecting the main concepts of the domain to be taught and its characteristics. This KB constitutes the domain ontology. As we are applying the CommonKADS methodology to formalize the knowledge acquisition process, the ontology will be represented by a set of concepts, attributes (describing the attributes), and expressions between this concepts.

Figure 7. Knowledge Acquisition for Domain Model Construction As shown in the Figure 7, a second module the Scheduler will generate the lesson plans, based on the acquired knowledge. These plans will be validated by the teacher. The knowledge acquisition module should also allow the update, by the teacher, of the ontology. We are now implementing the third and last component KC-Designer which is responsible for the knowledge contents and cycle of activities generation. It is also supported by the teacher. We are also addressing the problem of knowledge acquisition from multiple experts. In fact the obtained knowledge is more interesting since it involves multiple experts having different viewpoints. In previous works we have defined a CommonKADS extension for supporting multi-expertise [8]. We benefit especially from the associated developed techniques as conflict resolution.

7.

SHIECC IMPLEMENTATION

The SHIECC system is implemented in C++ Builder (a Visual Object Oriented Programming Language). The knowledge domain is implemented using the Toolbook 4.0

0-7803-6424-4/00/$10.00 2000 IEEE October 18 - 21, 2000 Kansas City, MO 30 th ASEE/IEEE Frontiers in Education Conference F1B-17

Session F1B
Development Kit. The implementation of the communication services is based on the TCP/IP protocols. All the modules of the system are integrated within a microcomputer Network made up of a server and several PCs clients. These microcomputers are related using an Intranet that we have already implemented.

REFERENCES
[1]. Antao, B. A, Brodersen, J., Bourne, J. R. and Cantwell, J.R.
Building Intelligent Tutorial Systems for Teaching Simulation in Engineering Education. IEEE Transactions on Education, V. 35, N. 1. February 1992.

[2]. Aronson, E. and others. The Jigsaw Classroom. Beverly Hills,


Calif.: Sage 1978.

8.

CONCLUSION

The SHIECC system proposed here is one of the little software packages that implement effective collaborative learning paradigm. It offers new opportunities in education by integrating collaborative learning with computer, multimedia, and network technologies in a manner that we believe will challenge traditional methods of pedagogy and benefit the learning process in a fundamental way. This proposal investigated a study, in the area of computer supported collaborative learning (CSCL), dealing with the problem of Knowledge Acquisition (AC) for semiautomatic construction of the domain model of an artificial tutor in a given discipline. Our approach proposes the definition and implementation of an intelligent module that, starting from the information supplied by the Specialist (the teacher of the discipline), come to build the Knowledge Base of the system (KB). A second module (the designer) will create the lessons, based on the acquired knowledge. In the area of engineering and mathematics [ [5], the proposed 1] model and its approach are very adequate and interesting. The compound student model enables to model both the students and their groups. The introduction of the concept of hypothesis gives the model another dimension, more efficient and realistic. As perspective of our work is the validation of the KCDesigner. Another important perspective is the integration of the fuzzy logic, in student modeling. This is more adapted for representing group uncertainty and vagueness.

[3]. Boy, G. A. Software Agents for Cooperative Learning. In Software


Agents. Edited Bradshaw J. M. AAAI Press/MIT Press. 1997.

[4]. Ellis S. S. and Whalen S. F. Cooperative Learnin g: Getting Started.


Scholastic, New York. 1990.

[5]. Hosotani, K. Japanese Quality Concept: An Overview. Quality


Resources. White Plains, NY, USA. 1992.

[6]. Johnson, D. W. and Johnson, T. J. Learning Together and Alone.


Cooperative, Competitive and Indiv idualistic Learning. Allyn and Bacon. 1994.

[7]. Labidi, S. CommonKADS Extension for Supporting MultiExpertise. In Proceedings of the 17th International Conference of the British Computer Society on Expert Systems, (ES'97). Churchill College, Cambridge. December 15-17, 1997. United Kingdom.

[8]. Labidi, S. and Ferreira, J. S. Technology -Assisted Instruction


Applied to Cooperative Learning: the SHIECC Project . In Proceedings of the IEEE International Conference Frontiers in Education (FIE98). Tempe, Arizona. November 4-7, 1998.

[9]. Labidi, S. Lima, C. M. and Sousa C.M. Modeling Agents and their
Interaction within SHIECC: A computer Supported Cooperative Learning framework. To appear in the International Journal of Continuous Engineering and Life-Long Learning. Special Issues on Intelligent Agents for Education and Training System. 2000.

[10].

Shaw T. W. Development of an electronic classroom: the promise, the possibilities and the practicalities. In Engineering Science and Education Journal. April, 1995. Slavin, R. E. Cooperative Learning: Theory, Research and Practice. Prentice Hall, 1990.

[11].

0-7803-6424-4/00/$10.00 2000 IEEE October 18 - 21, 2000 Kansas City, MO 30 th ASEE/IEEE Frontiers in Education Conference F1B-18

You might also like