Letters of recommendation help open gates to university admissions, jobs, and scholarships. Now that more colleges are ditching SAT/ACT requirements and medical schools are grading pass/fail, these letters will likely rise in importance. As a decision-maker myself, I rely on a letter writer’s personal judgment in vouching for (or not mentioning) an applicant’s work ethic, experience, and character.
At least I did—until the subject of one of my own letters added paragraphs of self-praise to the PDF above my signature. It turns out his other would-be letter writers had asked him to ghostwrite their letters, so I asked around. Students, residents, and people in the working world all told me that being asked to write their own letters was common practice.
Certainly, pawning off letter writing saved time that busy people didn’t have, and what better ghostwriter to ask than the person most interested in a great narrative. Yet as a reader, I felt duped—I had always assumed that the signer had composed the words on the page. As a plastic surgeon, I am required to disclose conflicts of interest when I give lectures and publish scientific articles, but there are no such standards for letters of recommendation.
A letter from a Nobel Prize winner is more likely to win over decision-makers than one from a teaching assistant, but if the laureate doesn’t really know the applicant, their endorsement seems deceptive. The process rewards students coached in whom to ask and those brimming with more chutzpah than excellence. The applicants who are handed the pen and can craft professorial adulation about themselves are likely to take home the prize.
Some institutions support ghostwritten letters because no one knows an applicant’s accomplishments better than the applicant themself. But self-written letters introduce self-bias, also known as confidence—sometimes, over-confidence. Here’s where women strikeout. In our culture, women are raised to be accommodating and concerned with the welfare of others. Men, on the other hand, are socialized to be more competitive and assertive. Google discloses no scholarly studies of recommendation letters secretly written by the applicant (who would admit to doing so?), but cultural differences weaken women’s negotiating skills, at least for themselves—they negotiate well for others. By extension, ghostwriting men may champion themselves better than ghostwriting women.
Even if we can effectively ban ghostwritten letters, authentic letters of recommendation still favor men and non-underrepresented minorities. A 2019 analysis of third-year medical students’ narrative academic evaluations found that faculty mentioned personal attributes more often for women (“lovely”) and underrepresented minorities (“pleasant”) than for men. In contrast, they were more likely to use competency-related terms for men (“scientific”). Because evaluators, like everyone, are unaware of their own biases, they are more likely to give women and minorities lower scores, even defying objective data. The authors found narrative evaluations to be so biased that they urged medical schools to train evaluators in de-biasing techniques or to adopt alternatives, such as task demonstrations and blinded evaluations.
In the business world, researchers Frank Dobbin and Alexandra Kalev came up with a reliable alternative to narrative performance reviews. They replaced open-ended questions about employees with a list of specific competencies paired with an important requirement—evaluators had to support their ratings with specific examples. The requirement to prove an assessment with evidence reined in implicit bias, and the scores of underrepresented minorities jumped.
The connection between bias in letters of recommendation and other narrative assessments may seem remote from patient outcomes, but medical schools and training programs select only those who submit the test scores, grades, and letters to recommend them. Small differences in assessment can have a large impact on grades, a phenomenon known as the “amplification cascade.” Adding a word here (“pleasant”) or leaving out a word there (“exceptional”) can disadvantage an otherwise good candidate. Tasking students to write their own letters adds an inexcusable layer of unreliability.
In the medical field, stripping bias from the evaluation process can improve patient care. Patients often seek doctors with whom they share race or gender. Minority physicians are more likely to serve minority communities and conduct minority-relevant research. Female surgical patients have better outcomes with female surgeons, and minority patients have better outcomes with minority doctors. Compared with the general population, female, Black, and Hispanic doctors are significantly underrepresented in my own field of surgery, and those in training leave at much greater rates than White or Asian men. With disproportionately few underrepresented minorities in medicine and few women in some medical fields, patients may search in vain for the doctor they want.
Recommendation letters are valuable only if they are reliable. It does no one any favor to recommend, hire, or promote students, residents, or employees who are not qualified, but we cannot ignore the insidious risk of bias in seemingly “neutral” evaluation tools.
Heather Furnas is an adjunct clinical associate professor at Stanford and an aesthetic surgeon in Northern California. She co-hosts the Skintuition podcast and can be reached on her website, Heather Furnas, and on Twitter @drheatherfurnas, Instagram @drheatherfurnas, Facebook, and LinkedIn.
