Design, Development and Evaluation of a Lightweight Knowledge-based System for Theoretically-grounded Math Error Classification

Mathematical problem solving is essential for developing analytical skills and critical thinking among learners. Traditional methods of detecting and correcting errors in mathematical problem-solving are manual and time-consuming, leading to delayed feedback, which can hinder effective learning. Intelligent Tutoring Systems (ITS) provide real-time, error-specific feedback but require direct interaction with computational devices, limiting their use in classrooms with insufficient technological infrastructure. ITS unplugged (ITS-U) have been proposed to address these technological barriers, but no research has investigated how to design a lightweight error classificatin systems aligned to the resource-constrained setting of ITS-U. Therefore, this paper presents the Mathematical Solutions Error Classification (MSEC) API, a lightweight, knowledge-based tool designed to automate the classification of errors in mathematical solutions. This API categorizes errors into theoretically-grounded types, such as syntactical mistakes, conceptual misunderstandings, and calculation errors. MSEC is particularly suited for low-tech educational environments, aligning with the principles of ITS-U. We implemented MSEC within an ITS-U and conducted a case study involving 49 students, demonstrating its practical applicability and positive contribution to teaching and learning. The study highlights MSECs efficiency, ability to provide real-time feedback, and adaptability to various educational contexts, offering a significant advancement in automated error detection for ITS-U.