FairPair: A Robust Evaluation of Biases in Language Models through Paired Perturbations

Jane Dwivedi-Yu


Abstract
The accurate evaluation of differential treatment in language models to specific groups is critical to ensuring a positive and safe user experience. An ideal evaluation should have the properties of being robust, extendable to new groups or attributes, and being able to capture biases that appear in typical usage (rather than just extreme, rare cases). Relatedly, bias evaluation should surface not only egregious biases but also ones that are subtle and commonplace, such as a likelihood for talking about appearances with regard to women. We present FairPair, an evaluation framework for assessing differential treatment that occurs during ordinary usage. FairPair operates through counterfactual pairs, but crucially, the paired continuations are grounded in the same demographic group, which ensures equivalent comparison. Additionally, unlike prior work, our method factors in the inherent variability that comes from the generation process itself by measuring the sampling variability. We present an evaluation of several commonly used generative models and a qualitative analysis that indicates a preference for discussing family and hobbies with regard to women.
Anthology ID:
2024.safety4convai-1.4
Volume:
Proceedings of Safety4ConvAI: The Third Workshop on Safety for Conversational AI @ LREC-COLING 2024
Month:
May
Year:
2024
Address:
Torino, Italia
Editors:
Tanvi Dinkar, Giuseppe Attanasio, Amanda Cercas Curry, Ioannis Konstas, Dirk Hovy, Verena Rieser
Venues:
Safety4ConvAI | WS
SIG:
Publisher:
ELRA and ICCL
Note:
Pages:
28–39
Language:
URL:
https://aclanthology.org/2024.safety4convai-1.4
DOI:
Bibkey:
Cite (ACL):
Jane Dwivedi-Yu. 2024. FairPair: A Robust Evaluation of Biases in Language Models through Paired Perturbations. In Proceedings of Safety4ConvAI: The Third Workshop on Safety for Conversational AI @ LREC-COLING 2024, pages 28–39, Torino, Italia. ELRA and ICCL.
Cite (Informal):
FairPair: A Robust Evaluation of Biases in Language Models through Paired Perturbations (Dwivedi-Yu, Safety4ConvAI-WS 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.safety4convai-1.4.pdf