Helping Teachers Help Their Students: A Human-AI Hybrid Approach

There is a global interest in artificial intelligence to support online learning, but little increase in support for online professors, teachers and tutors (instructors). Over time, more students join online learning, but instructors have no equivalent increase in support to manage their online classes, leaving students under-served. This is evidenced by the number of students who dropout or fail online courses, blaming the “lack of support” from instructors. Interactions in such courses generate considerable quantity and diversity of data, allowing the extraction of pedagogically relevant information. However, instructors do not master the techniques and technologies needed to do it, and it is not practical to train them to do so. In this work, we propose an authoring tool (called T-Partner) that implements a process we created to deal with educational data. The objective is to support instructors making informed pedagogical decisions to manage their online course. T-Partner promotes the cooperation between artificial and human intelligences, however we do not know the appropriate balance between these “intelligences”. We then created two versions of the T-Partner to help instructors to: (1) find relevant pedagogical situations occurring within their online courses; (2) understand these situations; (3) create interventions (study plans, for example) to address these situations; (4) monitor and evaluate the impact of these interventions. We evaluated if both versions allowed instructors to make pedagogical decisions and their perceptions regarding this support to decision-making. The results show that both versions brought benefits to pedagogical decision-making, and were positively perceived by the participants.