Harms of Gender Exclusivity and Challenges in Non-Binary Representation in Language Technologies

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 1968–1994 November 7–11, 2021. c⃝2021 Association for Computational Linguistics 1968 Harms of Gender Exclusivity and Challenges in Non-Binary Representation in Language Technologies Sunipa Dev she/her UCLA Masoud Monajatipoor* he/him UCLA Anaelia Ovalle* they/he/she UCLA Arjun Subramonian* they/them UCLA, Queer in AI Jeff M Phillips he/him University of Utah Kai-Wei Chang he/him UCLA Abstract Content Warning: This paper contains exam- ples of stereotypes and associations, misgen- dering, erasure, and other harms that could be offensive and triggering to trans and non- binary individuals. Gender is widely discussed in the context of language tasks and when examining the stereo- types propagated by language models. How- ever, current discussions primarily treat gen- der as binary, which can perpetuate harms such as the cyclical erasure of non-binary gender identities. These harms are driven by model and dataset biases, which are consequences of the non-recognition and lack of understand- ing of non-binary genders in society. In this paper, we explain the complexity of gender and language around it, and survey non-binary persons to understand harms associated with the treatment of gender as binary in English language technologies. We also detail how current language representations (e.g., GloVe, BERT) capture and perpetuate these harms and related challenges that need to be acknowl- edged and addressed for representations to eq- uitably encode gender information. 1 Introduction As language models are more prolifically used in language processing applications, ensuring a higher degree of fairness in associations made by their learned representations and intervening in any biased decisions they make has become in- creasingly important. Recent work analyzes, quan- tifies, and mitigates language model biases such as gender, race or religion-related stereotypes in static word embeddings (GloVe (Pennington et al., 2014)) and contextual (e.g., BERT (Devlin et al., 2019)) representations (Bolukbasi et al., 2016; De- Arteaga et al., 2019; Ravfogel et al., 2020; Dev et al., 2020b). * Equal contribution {sunipa,anaelia,arjunsub,kwchang}@cs.ucla.edu, monajati@g.ucla.edu, jeffp@cs.utah.edu A bulk of social bias studies on language mod- els have focused on binary gender and the stereo- types associated with masculine and feminine at- tributes (Bolukbasi et al., 2016; Webster et al., 2018; Dev et al., 2020b). Additionally, models of- ten rely on gendered information for decision mak- ing, such as in named entity recognition, corefer- ence resolution, and machine translation (Mehrabi et al., 2020; Zhao et al., 2018; Stanovsky et al., 2019), but the purview of gender in these tasks and associated measures of performance focus on bi- nary gender. While discussing binary gender bias and improving model performance are important, it is important to reshape our understanding of gen- der in language technologies in a more accurate, inclusive, non-binary manner. Current language models can perpetrate harms such as the cyclical erasure of non-binary gender identities (Uppunda et al., 2021; Sap, 2021; Lakoff; Fiske, 1993; Fast et al., 2016; Behm-Morawitz and Mastro, 2008). These harms are driven by model and dataset biases due to tainted examples, limited features, and sample size disparities …

Content truncated to 500 words