We survey 146 papers analyzing ``bias'' in NLP systems, finding that their motivations are often vague, inconsistent, and lacking in normative reasoning, despite the fact that analyzing ``bias'' is an inherently normative process. We further find that these papers' proposed quantitative techniques for measuring or mitigating ``bias'' are poorly matched to their motivations and do not engage with the relevant literature outside of NLP. Based on these findings, we describe the beginnings of a path forward by proposing three recommendations that should guide work analyzing ``bias'' in NLP systems. These recommendations rest on a greater recognition of the relationships between language and social hierarchies, encouraging researchers and practitioners to articulate their conceptualizations of ``bias''---i.e., what kinds of system behaviors are harmful, in what ways, to whom, and why, as well as the normative reasoning underlying these statements---and to center work around the lived experiences of members of communities affected by NLP systems, while interrogating and reimagining the power relations between technologists and such communities.
Description
Language (Technology) is Power: A Critical Survey of “Bias” in NLP - ACL Anthology
%0 Conference Paper
%1 blodgett-etal-2020-language
%A Blodgett, Su Lin
%A Barocas, Solon
%A Daumé III, Hal
%A Wallach, Hanna
%B Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics
%C Online
%D 2020
%I Association for Computational Linguistics
%K bias language lm models nlp survey
%P 5454--5476
%R 10.18653/v1/2020.acl-main.485
%T Language (Technology) is Power: A Critical Survey of ``Bias'' in NLP
%U https://www.aclweb.org/anthology/2020.acl-main.485
%X We survey 146 papers analyzing ``bias'' in NLP systems, finding that their motivations are often vague, inconsistent, and lacking in normative reasoning, despite the fact that analyzing ``bias'' is an inherently normative process. We further find that these papers' proposed quantitative techniques for measuring or mitigating ``bias'' are poorly matched to their motivations and do not engage with the relevant literature outside of NLP. Based on these findings, we describe the beginnings of a path forward by proposing three recommendations that should guide work analyzing ``bias'' in NLP systems. These recommendations rest on a greater recognition of the relationships between language and social hierarchies, encouraging researchers and practitioners to articulate their conceptualizations of ``bias''---i.e., what kinds of system behaviors are harmful, in what ways, to whom, and why, as well as the normative reasoning underlying these statements---and to center work around the lived experiences of members of communities affected by NLP systems, while interrogating and reimagining the power relations between technologists and such communities.
@inproceedings{blodgett-etal-2020-language,
abstract = {We survey 146 papers analyzing {``}bias{''} in NLP systems, finding that their motivations are often vague, inconsistent, and lacking in normative reasoning, despite the fact that analyzing {``}bias{''} is an inherently normative process. We further find that these papers{'} proposed quantitative techniques for measuring or mitigating {``}bias{''} are poorly matched to their motivations and do not engage with the relevant literature outside of NLP. Based on these findings, we describe the beginnings of a path forward by proposing three recommendations that should guide work analyzing {``}bias{''} in NLP systems. These recommendations rest on a greater recognition of the relationships between language and social hierarchies, encouraging researchers and practitioners to articulate their conceptualizations of {``}bias{''}---i.e., what kinds of system behaviors are harmful, in what ways, to whom, and why, as well as the normative reasoning underlying these statements{---}and to center work around the lived experiences of members of communities affected by NLP systems, while interrogating and reimagining the power relations between technologists and such communities.},
added-at = {2021-01-25T09:09:41.000+0100},
address = {Online},
author = {Blodgett, Su Lin and Barocas, Solon and Daum{\'e} III, Hal and Wallach, Hanna},
biburl = {https://www.bibsonomy.org/bibtex/219eac32b8c9e7cdad9cf9827e3f7ecd1/schwemmlein},
booktitle = {Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics},
description = {Language (Technology) is Power: A Critical Survey of “Bias” in NLP - ACL Anthology},
doi = {10.18653/v1/2020.acl-main.485},
interhash = {f6c0cc9c68e070af0f07f06ef23c4c48},
intrahash = {19eac32b8c9e7cdad9cf9827e3f7ecd1},
keywords = {bias language lm models nlp survey},
month = jul,
pages = {5454--5476},
publisher = {Association for Computational Linguistics},
timestamp = {2021-01-25T09:09:41.000+0100},
title = {Language (Technology) is Power: A Critical Survey of {``}Bias{''} in {NLP}},
url = {https://www.aclweb.org/anthology/2020.acl-main.485},
year = 2020
}