Building an Educational Recommender System based on Conceptual Change Learning Theory to Improve Students' Understanding of Science Concepts
Author(s)Okoye, Ifeyinwa Uchechukwu
educational recommender system
education computational models
natural language processing
Full recordShow full item record
AbstractScience misconceptions can be deeply held and difficult to change. Conceptual change learning theory (CCLT) applied in the classroom environment has been effective in helping students remedy their persistent science misconceptions. This work studies CCLT in an online learner-driven environment using five related studies. The first study creates the online learner-driven environment while the second study evaluates the resulting system. The last three studies create computational models that accomplish a teacher's task per CCLT. The first study uses participatory design methods to create the online learner-driven environment called CLICK2. This is an effective feedback environment because it can answer the questions: where am I? where am I going? how am I going? and where to next? in relation to a user's knowledge state. Results show that users were satisfied with the interaction design of CLICK2. The second study is a learning study that investigated how CLICK2 in- fluences learners' processes and outcomes. Results show that using CLICK2 improved users' understanding of seasons and their confidence in understanding this concept. The last three studies use techniques from machine learning and natural language processing to perform three critical tasks underpinning support for CCLT: prioritizing learners' misconceptions, extracting core concepts, i.e., learning goals and sequencing them. All three studies draw on analyses of human expert processes to inform the design and evaluation of the algorithms. Results show that an alignment of sequenced core concepts to misconceptions in a learner's work is a good feature for prioritizing misconceptions in the learner's work; reducing the extraction rate in a multi-document summarizer produces good core concepts; and dynamically generating useful pedagogical sequences from a list of core concepts is feasible. This work contributes to the scientific literature by introducing a methodology for automatically prioritizing core concepts and student misconceptions in a pedagogically useful manner. Furthermore, this work shows that conceptual change learning theory can be implemented in an online learner-driven environment.