Evaluating the quality of medical multiple‐choice items created with automated processes
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
OBJECTIVES: Computerised assessment raises formidable challenges because it requires large numbers of test items. Automatic item generation (AIG) can help address this test development problem because it yields large numbers of new items both quickly and efficiently. To date, however, the quality of the items produced using a generative approach has not been evaluated. The purpose of this study was to determine whether automatic processes yield items that meet standards of quality that are appropriate for medical testing. Quality was evaluated firstly by subjecting items created using both AIG and traditional processes to rating by a four-member expert medical panel using indicators of multiple-choice item quality, and secondly by asking the panellists to identify which items were developed using AIG in a blind review. METHODS: Fifteen items from the domain of therapeutics were created in three different experimental test development conditions. The first 15 items were created by content specialists using traditional test development methods (Group 1 Traditional). The second 15 items were created by the same content specialists using AIG methods (Group 1 AIG). The third 15 items were created by a new group of content specialists using traditional methods (Group 2 Traditional). These 45 items were then evaluated for quality by a four-member panel of medical experts and were subsequently categorised as either Traditional or AIG items. RESULTS: Three outcomes were reported: (i) the items produced using traditional and AIG processes were comparable on seven of eight indicators of multiple-choice item quality; (ii) AIG items can be differentiated from Traditional items by the quality of their distractors, and (iii) the overall predictive accuracy of the four expert medical panellists was 42%. CONCLUSIONS: Items generated by AIG methods are, for the most part, equivalent to traditionally developed items from the perspective of expert medical reviewers. While the AIG method produced comparatively fewer plausible distractors than the traditional method, medical experts cannot consistently distinguish AIG items from traditionally developed items in a blind review.
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.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.024 | 0.877 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.003 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.004 | 0.000 |
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 it