Selective Mixup for Debiasing Question Selection in Computerized Adaptive Testing

Computerized Adaptive Testing is a widely used technology for evaluating examinees' proficiency in online education platforms. By leveraging prior estimates of proficiency to select questions and updating the estimates iteratively based on responses, it enables personalized examinee modeling and has attracted substantial attention. Despite this progress, most existing works focus primarily on improving proficiency estimation accuracy, while overlooking the selection bias inherent in the adaptive process. Selection bias arises because the question selection is strongly influenced by the estimated proficiency, such as assigning easier questions to examinees with lower proficiency and harder ones to examinees with higher proficiency. Since the selection depends on prior estimation, this bias propagates into the diagnostic model, which is further amplified during iterative updates, leading to misaligned and biased predictions. Moreover, the imbalance in examinees' historical interactions often exacerbates bias in diagnostic models. To address this issue, we propose a debiasing framework consisting of two key modules: Cross-Attribute Examinee Retrieval and Selective Mixup-based Regularization. First, we retrieve balanced examinees with relatively even distributions of correct and incorrect responses and use them as neutral references for biased examinees. Then, Mixup is applied between each biased examinee and its matched balanced counterpart under label consistency. This augmentation enriches the diversity of bias-conflicting samples and smooths selection boundaries. Finally, extensive experiments on two benchmark datasets with multiple advanced diagnosis models have been conducted. The results demonstrate that our method substantially improves the generalization ability of question selection.