You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Implements a Delphi of GANs approach to detect bias in pre-trained language models and provides a novel way to assess and compare bias levels in different models across multiple dimensions such as gender, racial, and age bias
Code and data for the paper "An Information-Theoretic Approach and Dataset for Probing Gender Stereotypes in Multilingual Masked Language Models" (Findings of NAACL 2022)
Diving into the challenges of data quality in women's healthcare, exploring real-life stories that highlight the critical need for accurate data collection and its impact on diagnoses, treatment, and pharmaceutical advancements. Shedding light on the persistent issues of gender bias in healthcare and advocating for the importance of quality data.
A website that visualises how text embeddings trained on Google News are biased towards gender. It also functions as an open source database to generate text embeddings .
The project can be split into different sub-projects (easy difficulty: replication of the published meta-analysis for evidence of gender bias in hiring decisions; medium for newer modelling). Requires skills in R and will require some learning on Bayesian modelling.
Using GSS data, this paper tracks public perceptions of women in politics over time, in correlation with demographic factors, political views, and party identification