TY - GEN

T1 - Is the Lottery Fair? Evaluating Winning Tickets Across Demographics

AU - Hansen, Victor Petrén Bach

AU - Søgaard, Anders

N1 - Publisher Copyright: © 2021 Association for Computational Linguistics

PY - 2021

Y1 - 2021

N2 - Recent studies have suggested that weight pruning, e.g. using lottery ticket extraction techniques (Frankle and Carbin, 2018), comes at the risk of compromising the group fairness of machine learning models (Paganini, 2020; Hooker et al., 2020), but to the best of our knowledge, no one has empirically evaluated this hypothesis at scale in the context of natural language processing. We present experiments with two text classification datasets annotated with demographic information: the Trustpilot Corpus (sentiment) and CivilComments (toxicity). We evaluate the fairness of lottery ticket extraction through layer-wise and global weight pruning across three languages and two tasks. Our results suggest that there is a small increase in group disparity, which is most pronounced at high pruning rates and correlates with instability. The fairness of models trained with distributionally robust optimization objectives is sometimes less sensitive to pruning, but results are not consistent. The code for our experiments is available at https://github.com/vpetren/fairness_lottery.

AB - Recent studies have suggested that weight pruning, e.g. using lottery ticket extraction techniques (Frankle and Carbin, 2018), comes at the risk of compromising the group fairness of machine learning models (Paganini, 2020; Hooker et al., 2020), but to the best of our knowledge, no one has empirically evaluated this hypothesis at scale in the context of natural language processing. We present experiments with two text classification datasets annotated with demographic information: the Trustpilot Corpus (sentiment) and CivilComments (toxicity). We evaluate the fairness of lottery ticket extraction through layer-wise and global weight pruning across three languages and two tasks. Our results suggest that there is a small increase in group disparity, which is most pronounced at high pruning rates and correlates with instability. The fairness of models trained with distributionally robust optimization objectives is sometimes less sensitive to pruning, but results are not consistent. The code for our experiments is available at https://github.com/vpetren/fairness_lottery.

U2 - 10.18653/v1/2021.findings-acl.284

DO - 10.18653/v1/2021.findings-acl.284

M3 - Article in proceedings

AN - SCOPUS:85123915679

SP - 3214

EP - 3224

BT - Findings of the Association for Computational Linguistics

A2 - Zong, Chengqing

A2 - Xia, Fei

A2 - Li, Wenjie

A2 - Navigli, Roberto

PB - Association for Computational Linguistics

T2 - Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021

Y2 - 1 August 2021 through 6 August 2021

ER -