A method for identifying extreme OSCE examiners
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
BACKGROUND: Performance assessments rely on human judgment, and are vulnerable to rater effects (e.g. leniency or harshness). Making valid inferences from performance ratings for high-stakes decisions requires the management of rater effects. A simple method for detecting extreme raters that does not require sophisticated statistical knowledge or software has been developed as part of the quality assurance process for objective structured clinical examinations (OSCEs). We believe it is applicable to a range of examinations that rely on human raters. METHODS: The method has three steps. First, extreme raters are identified by comparing individual rater means with the mean of all raters. A rater is deemed extreme if their mean was three standard deviations below (hawks) or above (doves) the overall mean. This criterion is adjustable. Second, the distribution of an extreme rater's scores was compared with the overall distribution for the station. This step mitigates a station effect. Third, the cohort of candidates seen by the rater is examined to ensure that any cohort effect is ruled out. RESULTS AND IMPLICATIONS: Of 3000+ raters, fewer than 0.3% have been identified as being extreme using the proposed criteria. Rater performance is being monitored on a regular basis, and the impact of these raters on candidate results will be considered before results are finalised. Extreme raters are contacted by the organisation to review their rating style. If this intervention fails to modify the rater's scoring pattern, the rater is no longer invited back. As more data are collected the organisation will assess them to inform the development of approaches to improve extreme rater performance.
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 | 0.042 | 0.017 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| 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) | 0.003 | 0.002 |
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