‘You're certainly relatively competent’: assessor bias due to recent experiences
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
CONTEXT: A recent study has suggested that assessors judge performance comparatively rather than against fixed standards. Ratings assigned to borderline trainees were found to be biased by previously seen candidates' performances. We extended that programme of investigation by examining these effects across a range of performance levels. Furthermore, we investigated whether confidence in the rating assigned predicts susceptibility to manipulation and whether prompting consideration of typical performance lessens the influence of recent experience. METHODS: Consultant doctors were randomised to groups within an internet experiment. The descending performance group judged videos of Foundation Year 1 (F1; postgraduate Year 1) doctors in descending order of proficiency; the ascending performance group judged the same videos in ascending order. For all videos, participants rated: (i) trainee competence; (ii) rater confidence and (iii) percentage better (the percentage of other F1 doctors who would perform better on the same task). RESULTS: Overall, the descending performance group assigned lower scores than the ascending performance group (2.97 [95% confidence interval 2.73-3.20] versus 3.50 [95% confidence interval 3.25-3.74]; F(1,47) = 9.80, p = 0.003, d = 0.52). Pairwise comparisons showed differences were significant for good and borderline performances. The percentage better ratings showed a similar pattern (descending performance mean = 57.4 [95% confidence interval 52.5-62.3], ascending performance mean = 43.4 [95% confidence interval 38.4-48.5]; F(1, 46) = 16.0, p < 0.001, d = 0.67). Confidence ratings did not vary by level of performance and showed no relationship with the effect of group. DISCUSSION: Assessors' judgements showed contrast effects at both good and borderline performance levels. Findings suggest that assessors use normative rather than criterion-referenced decision making while judging, and that the norms referenced are weakly represented in memory and easily influenced. Confidence ratings suggested a lack of insight into this phenomenon. Raters' judgements could be importantly influenced in ways that are unfair to candidates.
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 | 0.001 | 0.045 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.133 | 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