Abstract:

The development of flexible, reusable Learning Objects (LOs) and their availability on Web repositories has become a cornerstone to promote advances in many educational fields and disciplines. But, as individual words cannot independently produce meaning, LOs in themselves are insufficient to fully accommodate the different learning and studying styles, strategies and preferences of the students. Thus, offering personalized curriculum designs and learning routes to individual students' needs+profiles is essential to support better initiatives and experiences in e-learning.

There are different research approaches for course and curriculum composition, such as adaptive and dynamic courseware generation. Intuitively, the former selects LOs from a given repository to ensure that a student completes all the activities that a teacher deems important, but in a way which is appropriate for the targeted individuals. The latter observes the student's progress during his/her interaction with a general on-line course, and dynamically adapts it according to the student's requirements. Clearly, this involves a separate vision: first to generate a "kind of tailored" course and second to execute, monitor and fully adapt such a course to the student's needs.

Here we face these two views in a joint way from an Artificial Intelligence planning perspective. Planning offers very appealing possibilities for the development of e-learning environments that bring the right content to the right person, but also at the right time and with the right resources, which is usually missing in traditional e-learning. Hence, the advantage of using intelligent planning techniques is twofold. Firstly, they bridge the gap between the purely e-learning necessities and its personalization, by making use of a flexible mapping of standard LOs and courses into standard planning domains, which are then solved by state-of-the-art planners and/or Case-Based Planning (CPB) methods. Secondly, they go beyond the traditional e-learning insights and give support not only to adaptation and LO sequencing, but also to time+resource constraints and multi-criteria optimization metrics, which are important in real environments. This raises a challenge for a successful integration with Learning Management Systems (LMSs) that facilitate the dynamic navigation of contents/LOs, monitor the students' progress when following their proposed learning routes, check whether some discrepancies (differences w.r.t. the expected state) appear and react to them to adapt the routes to the new necessities -by means of an adaptation process that does not ignore the original student's interests and tries to reuse the original route as much as possible.

In order to use our approach, LMSs need to be extended to provide capabilities for monitoring and checking the students' progress when executing learning routes. And it is not only a matter of plugging in a planner to find a route, but also to offer additional mechanisms for both teachers and students to manage courses and automated tools to find out discrepancies to adapt the route -note that CBP techniques do not require to compute a new route from scratch, as they allow us to adapt high quality learning routes already executed in the past. We have currently performed this integration with Moodle, but we also expect to create integration modules for other open source LMSs, such as dotLRN and Claroline, which are widely spread in the e-learning community.

Keywords:

e-learning, personalization of learning routes.