Proceedings of the workshop "Adaptive Systems
and User Modeling on the World Wide Web",
Sixth International Conference on User Modeling, Chia Laguna, Sardinia, 2-5 June 1997
Feedback and Adaptive Interaction for WWW-based courses
Yury Tsybenko, Viatcheslav Bykov
Department of Dialogue and Tutoring Systems
Glushkov Institute of Cybernetics
40, Prospect Academica Glushkova
Kiev, 252 022 Ukraine
tel.:380 44 267 6930, fax:380 44 266 1570
Abstract: Two approaches for interaction with a student within www-based distant course are considered in this paper. The first is asynchronous interaction in which tutor communicates with students by email. The second is adaptive on-line testing to control the studentís studying. Testing is provided by the server-side software via Common Gateway Interface (CGI) and interactive HTML forms.
Throughout the world there is a growing interest in using telecommunication technology to support teaching and learning. A number of courses are now being offered completely over the Internet. Courses are made up of instructional sequences involving learning materials curried out in various educational environments. During the last years using the WWW for educational goals also becomes important to many educational technologists involved into design of innovative learning environments. In educational settings Web serves as organised recourse for a collection of structured learning materials with a set of hyperlinks (local and remote) in it. These features are educationally useful, but appear to be so routing now. It is necessary to do a research of use WWW beyond this standard features. During the last years WWW has been extensively studied to determine its learning-support possibilities beyond that of offering a home page. Schneider and K. Block  distinguished four main levels of WWW use in Education:
This paper concerns levels 3 and 4 of the Web use in Education.
Assessment of a student's knowledge and tutor's feedback are necessary elements of education and training. Students cannot learn much by only browsing a hypertext. Learning must be active. There are two different approaches in providing a feedback and interaction with a student in distant education. The first approach is based on a role of a teacher. In some forms of distant education the teacher remains the central component in the course [B. Collis, 1995]. The difference from a face-to-face form is that the students are at a distance of a teacher. A teacher presents learning materials via Web and communicates with his students asynchronously via email, computer conference or by publishing teacher's instructions on the Web. Such an approach requires an additional tutors efforts on answering student's questions and evaluating student's work.
The another approach [E. Schwartz et. al. 1996], [F Linn, et. al., 1996] [Antchev, et. al., 1996] is integrating distant educational power of WWW with interactivity and intelligence of Intelligent Tutoring Systems (ITS). Traditional ITS technique aims to develop some ITS features that usually performed by human teacher in the classroom. This technique is based on a model of a student. Student model was developed to help intelligent tutor decide which exercise to give to a student, when to interrupt, what level of explanation to give, and so on. This technology is also may be useful for distance education, when human teacher cannot interact with a student directly. However, it is difficult to transfer directly the ITS technology in the Web. The traditional ITS teaching strategies impose some constraints on student's behavior: typically, the system identifies a student, and basing on the student's model guides the student during learning sessions. In contrast, a Web student can freely browse through the materials available, the system never knows the path by which the student has arrived in a hypermedia document.
The existing client-server standards on the WWW support the development of distributed interactive applications using standard HTTP clients and servers. The mechanism that makes this possible are the Common-Gateway Interface (CGI) and interactive HTML forms. Such a form transmitted from the HTTP daemon to the WWW browser should be sent back to the server when completed. Each form is associated with a particular CGI program, which is activated by the Web server when the form is received. The CGI program processes a student's inputs and generates instructional materials in the form of HTML document, which becomes available to the student.
The WWW-based course called "Communication and Information Technologies" (CIT) is currently developed at Glushkov Institute of Cybernetics, Kiev, Ukraine. This is comprehensive course on basic Internet technologies consisting of 30 lessons. Each lesson contains on-line tests, exercises and other activities that form interactive learning environment of the lesson. The two different types of interaction are implemented for the CIT course. The first is asynchronous interaction in which tutor communicate with students by email. This form of interaction is used for sending tutor's instructions, answering student's questions and monitoring collaborative work within a group of students. The second is adaptive on-line interaction provided by the server-side software. This form of interaction is used for evaluating the results of student's studying.
The interactive forms allow to offer a variety of assessment methods such as true/false questions multiple-choice questions, short-answer and simple essay questions. The forms are spread through the course materials on the Web in order to supplement the main presentation chunks of the course with interaction facilities. The lesson assignment is a set of tests that relate to the lesson topics and should be answered simultaneously. Each lesson contains from 5 to 7 different tests.
The CIT course learning materials consist of lecture notes divided into chapters and lessons. Each lesson is a hypertext containing interactive on-line forms such as tests and exercises. After browsing the lesson materials a student has to perform the lesson assignment. In a typical interactive session the computer will pose a problem and ask the student to fill in on-line form. After all forms have been completed the lesson assignment is submitted for evaluation. The student's answer is assessed by the server software and the student's scoring along with analysis of student's mistakes are returned to the student. To improve his or her current scoring a student is allowed to perform each assignment several times. This is also useful for a student's self assessment. Each testing attempt has been evaluated by CGI program, which processes the student's inputs. The CGI program also generates new version of the test for the next testing attempt.
In order to be able to tutor successfully a system requires detail information about a student. The necessary information includes a description of current state of student's knowledge, past dialog history, etc. This demand to form a model of a student. Student modelling is necessary part of advanced computer-based teaching technologies applied say, in distance education, when human teacher may not interact with a student directly. Students are different: they have different prior knowledge,different interests, different learning aptitudes and so on. Without a student model a computer-based tutoring system (or method) will perform in exactly the same way with all users, since there is no basis to determine otherwise. In CIT course we use very simple student model consisting of two levels: novice and expert. There are two different types of exercises designed for each level. At the initial step the system assumes all students are novices. If during three testing session a student demonstrates stable high scoring his/her status is changed from novice to expert. At expert level the student has to pass trough more complex tests. However, if at expert level a student demonstrates unsatisfactory scoring his status is changed to novice.
To perform comprehensive student's knowledge testing during CIT course delivery a system must identify a student every time she/he loaded into the system. Normally, WWW-based systems can identify only the IP-address sent by client's computer. However, students can use different PCs working with the course. To bypass above problem the students are registered formally for the CIT course. This provides some access constraints on the course materials and allows the system to follow the student's performance. The student's inputs are recorded by the server-side software. This helps a tutor to control the student's studying basing on past dialog history. Until a lesson assignment has not been completed with satisfactory scoring, it is no reason to assume that a student has learned the lesson materials.
There are a lot of recently developed innovations such ATM multimedia, virtual reality, interactive video that will enhance the learning power of the Web. Current research demonstrates the usefulness of adaptive interaction technique for WWW-based learning. The intelligent interactive applications will only compliment the above innovations by performing a routine work on assessment of a student's knowledge and monitoring her or his learning. In addition intelligent features of the Web is enhanced factor for student's motivation and persistence.
In the proposed approach comprehensive tests are implemented to diagnose a student's knowledge and skills. The special database records information about the student's interaction with the system. This information will allow an intelligent tutor (student model) to infer current state of student knowledge and to provide adaptive interaction with the student. The most interesting future work is to develop a consistent model for a Web student. From technical point of view the ability of server side software to perform nontrivial computation is essential. Therefore the idea of incorporating advanced student modeling software as a part of server side software sounds natural.
[Antchev K., et al., 1996] Antchev K., Luhtalahti M., Multisilta J., Pohjolainen S., Suomela K. (1996). A WWW Microword for Mathematics. Proc of Ed-TELECOM'96 pp. 5-10.
[Collis B., 1995] Collis B. (Ed.) (1995). On-line and distance learning. URL http://www.to.utwente.nl/ism/online95/campus/campus.html.
[Schneider D., Block K., 1995] Schneider D., Block K. (1995)The World-Wide Web in Education. URL http://tecfa.unige.ch/tecfa/tecfa-research/CMC/andrea95/andrea.html
[Lin F., et al., 1996] Lin F., Danielson R., Herrgot S. (1996). Adaptive Interaction through WWW. Proc of Ed-TELECOM'96 pp. 173-178.
[Self J., 1991] Self J. (1991) Formal Approaches to Student Modelling. Technical Report AI-59. Lancaster University, UK.
[Schwarz E., et al., 1996] Schwarz E., Brusilovsky P., Weber G., World-Wide Intelligent Textbooks. (1996). Proc of Ed-TELECOM'96 pp. 302-307.
[Wenger E., 1987] Wenger E. (1987) Artificial Intelligence and Tutoring Systems. Los Altos: Morgan-Kaufman.