Abstract:

The presence and use of software agents in computer-assisted teaching environments is not a novelty. For several years, we have assisted to the materialization of a significant number of quite interesting R&D efforts, which produced very concrete eLearning applications using software agents – artificial tutors. The results were quite diversified, ranging for tools for helping teachers and educators in a large variety of teaching and learning processes, such as simple reading aid programs or even as highly sophisticated programs for relating educational materials, or as adaptive tools for supporting the preparation and execution of student evaluation processes, among others. The artificial tutors elevate an educational system to a new level. since they allow for the provision of educational contents according to the profile of students, reflecting their expertise and knowledge, and personalizing the learning processes through the development of descriptive models based on the knowledge and skills of the student revealed with the actions and decisions that were taken during all the working sessions the student had with the tutor. The Leonard System emerged from an R&D initiative with the goal to provide an on-line educational platform especially oriented for supporting evaluation processes on undergraduate courses at our university. It was designed following a much-diversified set of practical heuristics acquired during several years of lecturing on different courses. The system use a rule-based inference engine that works over a knowledge base containing evaluation heuristics and methods, profiling data structures, and event lecturing sessions records, which provide to the system the ability to define in real-time what is the best evaluation path for determining current expertise and knowledge of students. Following students during evaluation sessions, which are often used for knowing the current status of the students in a selection of lecturing topics, the system establishes (or redefines) their profile, in order to infer what are the most learning critical aspects students face at the moment. Having this knowledge, it is possible to recommend the lecturing topics students need to work (to improve) or make more importance to those topics in future evaluation sessions. These characteristics allow the system to adapt evaluation plans for each student, in particular, and follow along time the evolution of their learning process, providing very specific analytical data, such as key performance indicators based on student learning paths and evaluation processes performance. The system is supported by a Web-based platform, which ensure the main services of the system, namely user credentials definition, evaluation processes methods, structures and statistics, knowledge base maintenance and inference rules storage. The system configuration and components were created based on the idea of a system for students studying anytime, anywhere. In this paper we will present and discuss the work carried out in the development of the Leonardo system, paying particular attention to all the aspects related to the definition, execution and analysis of its evaluation unit based of student profiles, approaching as well data storage and processing mechanism, and, finally, a brief demonstration of how an evaluation process is carried out.

Keywords:

Educational Environments, Teaching Support Systems, Intelligent Tutoring Systems, Evaluation Processes Customization.