Abstract:

The format for a traditional web-course is linear. Learners go through learning materials one by one in the order set by the teachers. Learning materials are followed by a test for measuring the learning results. Although this kind of environment is convenient to deploy, such model does not support the notion of self-directed learning in which students should be supported according to individual skill levels and personal learning goals. Large drop-out rates in web courses demonstrate that online learning has not reached its potential. There is a need for more personal and supportive learning platforms.

Intelligent and adaptive tutoring systems have been found to be useful in guiding students in effective learning strategies and making progress visible to the learner. However, most settings are usually built around a specific learning material and focus on learning within a single session. We believe that using data and algorithms, it is possible to expand this kind of approach to a tutoring system when studying any kind of learning material.

CLANEDtm (http://app.claned.com) is an open learning platform, which can be used by organisations (for deploying online courses) and free by learners (studying and publishing learning materials). The vision of CLANEDtm is to use validated measurement tools to enhance the learning process. Measuring activities and experiences enables the system to give feedback to the students. In addition, data is used to help the teacher to design effective settings for learning.

The goal of the current project is to build the first phase of an intelligent tutoring system; a recommendation engine and supportive structure for instructing in self-directed learning. The inputs for the system are based on personal goals, individual needs, user history and the learning paths of other users. The learner is encouraged to set goals (self-articulated or pre-given), prioritize between different goals and schedule learning accordingly. The technical challenge of the project is to make various user given goals comprehensible to the recommendation engine. Thereafter, we can suggest appropriate learning materials and effective strategies for pursuing these goals.

Our technical approach is to map learner specified goals into key concepts. Relations between these concepts are mapped to form a learning path towards completing a goal. Learning materials are given specific positions in the learning paths. The system will give learners content recommendations, which are based on both the relevance of each material given a goal to be learned and the relevance of the material given the overall progression. The latter relevance is increased when the text belongs to
similar local neighbourhoods in past learning paths. For example, if two documents explain a particular mathematical concept well, only one may be well-suited for the learner when progressing in her learning path. Further, if a learner is faced with a document, which she lacks the background to understand, document matching using these structures allows the recommendation engine to trace prerequisite concepts and documents.

The system compares students’ learning activities encoded as multidimensional time series data to their eventual outcomes. The progress towards the goals is tracked, which allows the learners to be alerted in case of apparent problems such as falling behind the schedule or contradictions between claimed priorities and actual activities.

Keywords:

Intelligent Tutoring Systems, Web Based Instruction.