MCQ Difficulty Prediction via Modeling Learner Heterogeneity Using Data-Driven Cognitive Profiling

In three linesMCQ difficulty prediction via data-driven cognitive profiling. Persona-driven framework using latent class analysis (LCA) on EEDI dataset, LLM simulation of response distributions per persona, aggregation with topic context and Ridge Regression. Improvement: MSE 0.367→0.274, R²=0.686.

Summary generated by Claude — human-verified