Race and Gender Bias in Internal Medicine Program Director Letters of Recommendation
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: While program director (PD) letters of recommendation (LOR) are subject to bias, especially against those underrepresented in medicine, these letters are one of the most important factors in fellowship selection. Bias manifests in LOR in a number of ways, including biased use of agentic and communal terms, doubt raising language, and description of career trajectory. To reduce bias, specialty organizations have recommended standardized PD LOR. OBJECTIVE: This study examined PD LOR for applicants to a cardiology fellowship program to determine the mechanism of how bias is expressed and whether the 2017 Alliance for Academic Internal Medicine (AAIM) guidelines reduce bias. METHODS: Fifty-six LOR from applicants selected to interview at a cardiology fellowship during the 2019 and 2020 application cycles were selected using convenience sampling. LOR for underrepresented (Black, Latinx, women) and non-underrepresented applicants were analyzed using directed qualitative content analysis. Two coders used an iteratively refined codebook to code the transcripts. Data were analyzed using outputs from these codes, analytical memos were maintained, and themes summarized. RESULTS: With AAIM guidelines, there appeared to be reduced use of communal language for underrepresented applicants, which may represent less bias. However, in both LOR adherent and not adherent to the guidelines, underrepresented applicants were still more likely to be described using communal language, doubt raising language, and career trajectory bias. CONCLUSIONS: PDs used language in a biased way to describe underrepresented applicants in LOR. The AAIM guidelines reduced but did not eliminate this bias. We provide recommendations to PDs and the AAIM on how to continue to work to reduce this bias.
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 | 0.004 | 0.006 |
| 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 | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.001 | 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