Can Question Validation Criteria Improve the Quality of LLM-Generated Multiple-Choice Questions?
Bibliographic record
Abstract
Adaptive assessment has been extensively researched as a key strategy for providing formative assessments in education. Fundamental to adaptive assessment is the development and ongoing maintenance of a substantial and well-structured question bank. However, manual question development by human instructors is time-consuming and poses a significant barrier to adopting adaptive assessment. To address this problem, research has been conducted in the use of Large Language Models (LLMs) for automated question generation. However, LLMs tend to hallucinate, necessitating human-in-the-loop question validation. Furthermore, to ensure alignment with the instructional standards and learning outcomes of a course, it is essential that instructors verify that the generated questions satisfy a defined set of Question Validation Criteria (QVC). In this study, we evaluate the effectiveness of embedding QVC directly into LLM prompts for the automatic generation of multiple-choice questions (MCQ) with feedback for use in adaptive formative assessments. Using a novel web-based application for question generation, MCQs were produced for three undergraduate-level courses. Validation by subject matter experts revealed no statistically significant improvement in quality between questions generated with and without embedded QVC in the prompts. These findings contribute to the growing body of research indicating that prompt engineering alone may be insufficient to achieve improvements in the quality of generated output for educational purposes.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.095 | 0.485 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.003 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.003 | 0.005 |
| Open science | 0.002 | 0.003 |
| Research integrity | 0.002 | 0.001 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".