272 — Exploring the Color of Glass: Letters of Recommendation for Female and Male Medical Faculty

I recently started contributing to the Reading for Gender Bias GitHub repository that aims to lint recommendation letters for common gender-bias shortcomings. It’s an awesome community of contributors, and you should absolutely come get involved, no matter your programming abilities!

I decided to start reading the first reference listed in that README, which is this one, a 2003 paper by Trix & Psenka.

The paper exposes a set of common pitfalls recommenders use, rather intentionally or unintentionally, to “gatekeep” academia. In particular, they looked at three years worth of recommendation letters to a large American medical school. In the mid-1990s, when these letters originated, women’s chances of receiving tenure in the medical world was half that of male colleagues.

There was nothing outwardly nefarious about the content of the recommendation letters for women. But there were patterns in what, by failing to say, the authors were indeed saying.

The authors found several patterns that demonstrated most saliently the difference between letters written for men and letters written for women:

Length. Letter for women were notably shorter (227 words instead of 253 words). (Interestingly, the longest letter — which was written for a male student — was written by a woman.)

Minimal assurance. These phrases are phrases that barely count as recommendation, and instead come across as impersonal and unfocused. They suggest that this person could just barely meet the needs of their program.

Gendered terms. That’s all I’m going to say about that!

Negative language. Faint praise, hedges, irrelevancies, and doubt raisers occurred in 24% of letters for women, versus only 12% for men.

Grindstone adjectives. These adjectives — ‘hardworking’, ‘conscientious’, ‘dependable’, ‘meticulous’, ‘thorough’, ‘diligent’, ‘dedicated’, and ‘careful’ — occurred in 34% of letters for women, versus 23% of letters for men. (If you want to help us write a detector for these words, check out this GitHub issue!)

Repetition. Letters for men were far more likely to repeat positive words than letters for women.

Portrayal. Women are more commonly referred to as students and teachers, whereas men are referred to as researchers and professionals.

References to personal life. Letters for women called out to personal matters more than often those for men.

References to CV & publications. Letters for women spoke about their CV or publications less than those for men.

This list isn’t exhaustive.