On the Independence of Association Bias and Empirical Fairness in Language Models
Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review
Standard
On the Independence of Association Bias and Empirical Fairness in Language Models. / Cabello, Laura; Jørgensen, Anna Katrine; Søgaard, Anders.
Proceedings of the 6th ACM Conference on Fairness, Accountability, and Transparency, FAccT 2023. Association for Computing Machinery, Inc., 2023. p. 370-378.Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review
Harvard
APA
Vancouver
Author
Bibtex
}
RIS
TY - GEN
T1 - On the Independence of Association Bias and Empirical Fairness in Language Models
AU - Cabello, Laura
AU - Jørgensen, Anna Katrine
AU - Søgaard, Anders
N1 - Publisher Copyright: © 2023 ACM.
PY - 2023
Y1 - 2023
N2 - The societal impact of pre-trained language models has prompted researchers to probe them for strong associations between protected attributes and value-loaded terms, from slur to prestigious job titles. Such work is said to probe models for bias or fairness - or such probes 'into representational biases' are said to be 'motivated by fairness' - suggesting an intimate connection between bias and fairness. We provide conceptual clarity by distinguishing between association biases [11] and empirical fairness [56] and show the two can be independent. Our main contribution, however, is showing why this should not come as a surprise. To this end, we first provide a thought experiment, showing how association bias and empirical fairness can be completely orthogonal. Next, we provide empirical evidence that there is no correlation between bias metrics and fairness metrics across the most widely used language models. Finally, we survey the sociological and psychological literature and show how this literature provides ample support for expecting these metrics to be uncorrelated.
AB - The societal impact of pre-trained language models has prompted researchers to probe them for strong associations between protected attributes and value-loaded terms, from slur to prestigious job titles. Such work is said to probe models for bias or fairness - or such probes 'into representational biases' are said to be 'motivated by fairness' - suggesting an intimate connection between bias and fairness. We provide conceptual clarity by distinguishing between association biases [11] and empirical fairness [56] and show the two can be independent. Our main contribution, however, is showing why this should not come as a surprise. To this end, we first provide a thought experiment, showing how association bias and empirical fairness can be completely orthogonal. Next, we provide empirical evidence that there is no correlation between bias metrics and fairness metrics across the most widely used language models. Finally, we survey the sociological and psychological literature and show how this literature provides ample support for expecting these metrics to be uncorrelated.
KW - Fairness
KW - Natural Language Processing
KW - Representational Bias
U2 - 10.1145/3593013.3594004
DO - 10.1145/3593013.3594004
M3 - Article in proceedings
AN - SCOPUS:85163629760
SP - 370
EP - 378
BT - Proceedings of the 6th ACM Conference on Fairness, Accountability, and Transparency, FAccT 2023
PB - Association for Computing Machinery, Inc.
T2 - 6th ACM Conference on Fairness, Accountability, and Transparency, FAccT 2023
Y2 - 12 June 2023 through 15 June 2023
ER -
ID: 381563506