Develop an Explainable, Fair, and Interactive Knowledge Tracing System

Personalized recommendation systems have been widely utilized to introduce adaptability in learning. However, the objective of a truly fair recommendation system is not only to consider the capability of delivering the suitable learning material to the learner based on their course-specific needs, but also to effectively evaluate the learner’s requirement at that time based on the prerequisite topics or present performances in related courses in parallel. Additionally, there is another significant difference between existing AI-powered recommender systems in education and the desired interpretable and interactive AI-powered optimized personalized education system, that may truly enable a transformative pedagogical approach by explaining its decision and updating based on students’ feedback in real time. Furthermore, although the objective of a personalized pedagogical approach is to ensure equal access to materials, so that each student gets the best instruction and practice at the pace they need it and eventually all of them be able to master the content at hand, in practice, the personalized instructional support systems sometimes fall short of the entire fairness standard. Typically, the lower-performing students end up receiving less practice and instruction than they need. Thus, the existing educational technologies, which aim to benefit all learners, might disproportionately benefit more advantaged groups of learners.

In this project, we aim to develop a continuous, fair, interactive, and interpretable knowledge tracing system to monitor students’ evolving knowledge state during the learning process and accurately predict their performance on future exercises, wherein the model’s interpretability capacity would address typical students’ query that usually care more about why a specific item is recommended rather than which/what item is recommended. The proposed model would take advantage of students’ feedback on their proactive understanding about their individual knowledge states to improve the relevance and fairness of the recommendation decision.

• Research publication
• Prototype

Computer Science & Engineering