Towards Helping Teachers Select Optimal Content for Students

In a personalized learning context, teachers decide which content to assign to students on the basis of data. However, it is not clear that simply providing teachers with data is sufficient to promote good instructional decisions. In this paper, we study data from an online learning platform that gives teachers data on student test performance and then allows them to decide which new skill students should work on. We then apply a knowledge graph algorithm to infer whether the content the teacher assigned the student is a skill that the student is ready to learn (i.e. the skill is within the student’s Zone of Proximal Development), whether the student is not yet ready to learn the skill, or whether the student has already learned the skill. In this paper, we study how the teacher’s decision of what skills or topics the student should work on correlate to the student’s learning outcomes. We study this issue using logistic regression to compare whether students master more skills based on whether they are assigned ready-to-learn skills or unready-to-learn skills according to the knowledge graph. The results demonstrate that in both mathematics and English learning contexts, if the teacher selects skills which the student is assessed by the algorithm to be ready to learn, the student gains more mastery than if he or she is assigned skills he or she is not ready to learn. We conclude by proposing a visualization that more clearly surfaces the knowledge graph predictions to teachers.