CE:Knowledge Tracking Model Uncertainty Assessment Method Under Trusted Artificial Intelligence

Trustworthy AI is becoming a new focus in the field of artificial intelligence, and the study of its trustworthiness criteria is crucial for enhancing the trustworthiness of AI. Despite the plethora of methods that have been explored to assess the trustworthiness of AI, there remains a dearth of simple and intuitive assessment approaches. This paper focuses on the domain of knowledge tracing, combining various clustering techniques to propose a trustworthy evaluation method based on uncertainty computation. Through validation using clustering, Monte Carlo methods, and correlation analysis, our approach effectively examines the trustworthiness of multiple prominent knowledge tracing models. Employing open data sets and virtual data sets grounded in Item Response Theory (IRT), our method achieves commendable performance at a low cost, thereby providing a reference for research into the trustworthiness of knowledge tracing models. Finally, we summarize model errors and areas for improvement, and offer perspectives on scale and integrated trustworthy development. Experimental results demonstrate that our method achieved an accuracy rate of 81.04%.