Researchers have developed a new method for automatically estimating the difficulty of multiple-choice questions by explicitly modeling distractors as distinct components. This structure-aware approach, tested on Chilean datasets, significantly improved prediction accuracy compared to models that only considered the question stem and correct answer. The best performing architecture achieved R^2 scores of 0.83 for Natural Sciences and 0.71 for Social Sciences, demonstrating the value of structural information in distractors for educational assessment. AI
IMPACT Enhances automated educational assessment tools by providing more accurate difficulty estimates for multiple-choice questions.
RANK_REASON Academic paper detailing a new methodology for automatic question difficulty estimation. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →