NLP Methods May Actually Be Better Than Professors at Estimating Question Difficulty
This addresses the problem of inaccurate exam question difficulty estimation by professors, offering an incremental improvement using LLMs to enhance assessment quality in educational settings.
The study compared Large Language Model-based methods with professors in estimating True/False exam question difficulty for Neural Networks and Machine Learning courses, finding that directly asking Gemini 2.5 outperformed professors, and supervised learning using LLM uncertainties with only 42 training samples yielded even better results.
Estimating the difficulty of exam questions is essential for developing good exams, but professors are not always good at this task. We compare various Large Language Model-based methods with three professors in their ability to estimate what percentage of students will give correct answers on True/False exam questions in the areas of Neural Networks and Machine Learning. Our results show that the professors have limited ability to distinguish between easy and difficult questions and that they are outperformed by directly asking Gemini 2.5 to solve this task. Yet, we obtained even better results using uncertainties of the LLMs solving the questions in a supervised learning setting, using only 42 training samples. We conclude that supervised learning using LLM uncertainty can help professors better estimate the difficulty of exam questions, improving the quality of assessment.