Bias Correct

This program identifies bias related to gender, race, and ethnicity in writing samples such as evaluations and letters of recommendations. The program is currently rule-based and highlights common forms of bias described in the scientific literature.

Our goal is to help you write strong letters, statements, or evaluations by giving you a way to identify the implicit bias that shows up in everyone’s writing.

The program highlights potential bias and describes the associated rules, so you can understand the best practices and make changes to reduce bias. The rules are designed to be over-inclusive, which means they may highlight words or phrases that do not represent bias (e.g., “She is patient” may represent bias related to gender, but “Her patient notes were clear and well-organized” does not represent bias).

Gender, racial, and ethnic bias in evaluations and letters of recommendations often shows up in coded language (e.g., “articulate,” “competent”) or failure to mention concrete accomplishments. Addressing bias may mean adding examples of concrete accomplishments or removing coded language.

We are not trying to make all writing the same. For example, describing emotional intelligence often includes communal language (“patient,” “compassionate”) and is important, even if it is associated with gender bias. To address this type of bias, you can check that you are including accomplishments, superlatives, and agentic language (“independent,” “assertive,” “decisive”) as well. Ask yourself if you use the same words when writing for all genders. You might need to add communal language for some genders and not others.