Empowering Instructors with Collective Intelligence: Learning Problem-Solving Paths to Facilitate Feedback Generation
Thyago Tenório, Luiz Rodrigues, Seiji Isotani, Ig Ibert Bittencourt · 2023
The feedback on the Intelligent Tutoring Systems (ITS) providing the necessary support and guidance to students successfully complete a given task, improving their learning. However, building proper feedback demands time, whereas instructors are often overloaded, and several interactions with experts of the domain knowledge to know the several paths students, which is often infeasible and increases the costs and the complexity to develop ITSs at scale. To address this problem, we proposed a novel approach to build feedback using students’ collective intelligence (CI), where our ITS might learn such paths incrementally building a knowledge graph with more detail than those predicted by instructors, optimizing the identification of problem-solving paths, and facilitating the iterative designing of meaningful and fine-grained feedback for the instructors by presenting domain model’s updated, meaningful visualizations. To evaluate our approach we developed an ITS in a domain of numerical expressions that was used by 99 students. As a result, we observed that our approach helps to create a knowledge graph with a quality equivalent to that built by specialists in less time and considerably reducing the instructor’s overload.