Tiffany Y. Tang and Gordon I. McCalla
In this paper we discuss the mechanism of a recommender system recommending papers for an evolving web-based learning system. Our system is unique in three aspects. The first is that our learning environment can evolve based on the system’s observance of learners and their behaviors. Therefore, the fittest papers will survive the natural selections by learners: papers liked by learners will survive. The second is that we introduce a pedagogically layered similarity between items that have been read by learners and candidate items for recommendation, which is different and desirable, since we argue that papers that match a learner’s interest might not be pedagogically suitable for him/her. The third significance is that we propose to annotate each paper with temporal sequences of learners’ learning behaviors. By doing it, we can maintain the objectivity as well as integrity of the papers. In addition the accumulated sequences of learners can play a key role for a deeper understanding of their knowledge levels/states, which, in turn, provide "just-in-time" recommendations to support and encourage e-learning.