arXiv:2112.08910cs.CL
Degendering Resumes for Fair Algorithmic Resume Screening
Abstract
We investigate whether it is feasible to remove gendered information from resumes to mitigate potential bias in algorithmic resume screening. Using a corpus of 709k resumes from IT firms, we first train a series of models to classify the self-reported gender of the applicant, thereby measuring the extent and nature of gendered information encoded in resumes. We then conduct a series of gender obfuscation experiments, where we iteratively remove gendered information from resumes. Finally, we train a resume screening algorithm and investigate the trade-off between gender obfuscation and screening algorithm performance. Results show: (1) There is a significant amount of gendered information in resumes. (2) Lexicon-based gender obfuscation method (i.e. removing tokens that are predictive of gender) can reduce the amount of gendered information to a large extent. However, after a certain point, the performance of the resume screening algorithm starts suffering. (3) General-purpose gender debiasing methods for NLP models such as removing gender subspace from embeddings are not effective in obfuscating gender.
Link to the actual paper here: https://arxiv.org/pdf/2112.08910.pdf Edit after having read it: It's short and easy to read, I would suggest reading it instead of the article linked. I think that more work should be needed before making suggestions. For example, is there a…
ZababaView on HNParent context
Original paper: https://arxiv.org/pdf/2112.08910.pdf The entire premise of the research relies on the fact that the resume samples were well matched: in other words, the quality of male candidates was equal to that of female counterparts. In the paper, it says Specifically, we…
quantoView on HNParent context