Intelligent planning for curriculum personalization of e-Learning contents. Moodle as a case study

Abstract

The use and enhancement of technology in the field of (distance) education has increased the adoption of e-learning. Nowadays, a simple connection to the Internet makes it possible to access any on-line content. Thus, e-learning platforms based on Learning Management Systems (LMSs) permits us to eliminate the barriers of time and space that usually limit traditional teaching. LMSs facilitate the administration, storage, delivery, visualization and monitoring of e-learning contents to both students and teachers in a simple and functional way. There are many LMSs, such as Moodle, Docebo, Atutor, Ilias, .LRN, etc., widely used to support learning/teaching activities. For the best use of a LMS, it is fundamental not to consider it in an isolated way as a simple repository of learning contents, but as part of a larger system in which learning contents are aggregated for the construction of courses that can be fully personalized or adapted to the specific needs and abilities of each student. Curriculum personalization is faced in several ways by using several techniques, such as evolutionary algorithms, data mining techniques, decision support systems, etc., with the main objective to offer the best content to the most adequate person. In this thesis, we focus on Artificial Intelligence planning as a valuable formalism to describe actions (learning contents) in terms of preconditions (precedence relationships) and causal effects to find plans (learning paths) that entirely fit the students’ profiles. Thus, the integration of intelligent planning techniques into LMSs makes it possible to achieve the greatest learning benefits thanks to the automatic generation of customized learning paths. In particular, we focus on Moodle (Module Object-Oriented Dynamic Learning Environment) as a free, open-source PHP web application for producing modular Internet-based courses. Teachers and students interact in Moodle by means of activities (e.g. lessons, chats, SCORMs, forum, wikis, etc.) On the other hand, resources (e.g. text pages, web pages, links,etc.) are optionally used to transmit additional information regarding the activities. Consequently, courses can be created by appropriately combining activities and resources to deliver learning contents to the students. Moodle is flexible enough to model from small to big courses and it can be easily used and modified in order to extend its functionalities. The integration of planning in Moodle is not a straightforward task because Moodle and, in general, any LMS presents some limitations for this. The lesson is the most adequate Moodle’s activity to define causal and precedence relationships. But the lessons’ number and order must be fixed by the teacher. In other words, the execution sequence, i.e. learning path, may not take into consideration all the students’ needs, as determined by their profiles (background, learning style and goals). Thus, we have faced and solved the main limitations in order to integrate our intelligent planning approach in Moodle. First, complex relationships that usually appear when planning very elaborate courses cannot be easily defined in the form of (Moodle) lessons. We have overcome this limitation by using intermediate dummy lessons that simulate virtual transitions between learning states. Second, the information that students input into the platform are not always sufficient to exhaustively define their background and learning goals, indispensable to define customized learning paths. Again, our solution implies the creation of dummy lessons that help specify initial/goal learning states. Third, compiling a planning model from Moodle’s database is not intuitive because course properties are not easily available and, thus, we defined a detailed mapping. Fourth, once generated a plan by using a standard planner, a challenge arises to personalize the visualization and navigation of individual paths to each student. We coped with this by associating personal views to dynamically created groups of students. This work offers a scalability evaluation to demonstrate the applicability of the proposed approach. In particular, by using two standard planners it is possible to solve a number of tests involving courses of different sizes (up to 64 lessons) and with different numbers of students (up to 500 students). Courses containing up to 64 lessons and with 300 students are customized in less than 900 seconds that is an excellent result. From a more practical perspective, we have include a real demonstration of the functioning of the presented approach by implementing and customizing in Moodle a real Physics course of the University of Calabria (Italy). In conclusion, this thesis presents an AI planning approach to generate personalized learning paths within Moodle. In particular, it is explained how to generate and visualize personalized learning paths within Moodle. It is im portant to observe that in the presented system can use any standard planner and that the entire approach can be easily extrapolated to any other standard LMS. Furthermore, thanks to the flexibility of the approach, it is possible to adapt and customize our ideas to any type of learning course in any type of LMS