A Framework of Interactive Educational Measurement through Traditional Scale Transformation with LLM Agent

Traditional assessment scales and expert evaluation play a crucial role in education for student assessment and future development. Meanwhile, traditional scale and expert evaluation face the challenge of lacking active participation from students, and students may answer hastily. The results of the scale are easily affected by central tendency bias, compliance bias, and social expectation bias, which may affect the accuracy of the scale measurement. The expert consultation method consumes professional resources and cannot evaluate large-scale student populations. In this paper, we propose EduCA (Educational Contextualized Agents) for automatic student assessment. EduCA transforms traditional scales into interactive contextualized questions. This framework includes three modules, student portrait agent, scenarioized question generator, and measure controller, which are designed to alleviate bias in assessment and increase student engagement. We also propose an assessment system for further research.