Investigating Gender Bias with LLMs

During this six-month lab rotation, we explored how LLMs handle gender biases in academic program descriptions. Given the persistence of gender imbalances in various fields, the project examines how LLMs generate and interpret gender oriented language, judging whether their paraphrases align with traditional psychological measures of gender targetedness of a text.

The project highlights the risk of LLMs unintentionally reinforcing gender biases when generating program or generating job descriptions.

While much existing research focuses on increasing female representation in male-dominated fields, our study examines both sides of the equation. We also look into the under-explored area of encouraging more men to enter female-dominated professions, such as therapy and education.