Key points are not available for this paper at this time.
Abstract Equity in engineering education hinges on the ability to fairly evaluate students. A critical issue in assessment is whether cognitive fatigue, which is marked by decreased performance during prolonged cognitive tasks, is a component of student proficiency or should be considered a measurement error. This raises crucial concerns about assessment fairness and the potential risk of inequitable attrition rates in engineering. Cognitive fatigue is often related to individual fatigue thresholds. Therefore, an important question is how instructors can arrange challenging questions in an exam and select the optimal number of questions for assessments. Item Response Theory (IRT) elucidates the relationship between students' latent traits and item features. At the same time, Machine Learning (ML) predicts the impact of features like item difficulty, discrimination, item order, and response time on each student's response pattern. By combining IRT and ML models, this study shows the effect of cognitive fatigue on student responses and provides an AI tool to predict the cognitive fatigue threshold during the exams for each student. The result of this cognitive fatigue threshold prediction not only helps instructors reduce the impact of cognitive fatigue on student performance by making better decisions on their exam duration and distribution of difficult items but can also reduce cheating on exams. This study underscores the significance of fostering awareness among students and professors regarding cognitive states during assessments and the provision of constructive feedback on student performance.
Mehrabi et al. (Sun,) studied this question.