We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.CL

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computation and Language

Title: Intersectional Bias in Causal Language Models

Abstract: To examine whether intersectional bias can be observed in language generation, we examine \emph{GPT-2} and \emph{GPT-NEO} models, ranging in size from 124 million to ~2.7 billion parameters. We conduct an experiment combining up to three social categories - gender, religion and disability - into unconditional or zero-shot prompts used to generate sentences that are then analysed for sentiment. Our results confirm earlier tests conducted with auto-regressive causal models, including the \emph{GPT} family of models. We also illustrate why bias may be resistant to techniques that target single categories (e.g. gender, religion and race), as it can also manifest, in often subtle ways, in texts prompted by concatenated social categories. To address these difficulties, we suggest technical and community-based approaches need to combine to acknowledge and address complex and intersectional language model bias.
Comments: 18 pages, 4 figures
Subjects: Computation and Language (cs.CL); Machine Learning (cs.LG)
Cite as: arXiv:2107.07691 [cs.CL]
  (or arXiv:2107.07691v1 [cs.CL] for this version)

Submission history

From: Liam Magee [view email]
[v1] Fri, 16 Jul 2021 03:46:08 GMT (967kb,D)

Link back to: arXiv, form interface, contact.