A Survey of Race, Racism, and Anti-Racism in NLP

By Anjalie Field, Su Lin Blodgett, Zeerak Talat, Yulia Tsvetkov. 2021

151 citations recorded by the source corpus.

Despite inextricable ties between race and language, little work has considered race in NLP research and development. In this work, we survey 79 papers from the ACL anthology that mention race. These papers reveal various types of race-related bias in all stages of NLP model development, highlighting the need for proactive consideration of how NLP systems can uphold racial hierarchies. However, persistent gaps in research on race and NLP remain: race has been siloed as a niche topic and remains ignored in many NLP tasks; most work operationalizes race as a fixed single-dimensional variable with a ground-truth label, which risks reinforcing differences produced by historical racism; and the voices of historically marginalized people are nearly absent in NLP literature. By identifying where and how NLP literature has and has not considered race, especially in comparison to related fields, our work calls for inclusion and racial justice in NLP research practices.

Subfields: Artificial Intelligence, Communication, Natural Language Processing, Safety, Robustness & Interpretability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Communication

In Natural Language Processing

In Safety, Robustness & Interpretability

Citations in the corpus

Language Generation Models Can Cause Harm: So What Can We Do About It? An Actionable Survey (2022, positive): “Evidence for this behavior has been substantially corroborated in NLP literature (Nadeem et al., 2021; Nozza et al., 2021; Huang et al., 2020b; Hassan et al., 2021; Blodgett et al., 2020; Liang et al., 2021; Field et al., 2021, inter alia).”
Hi, my name is Martha: Using names to measure and mitigate bias in generative dialogue models (2021, background): “Furthermore, our reliance on single-dimensional, US-Census-derived race/ethnicity categories can have the effect of marginalizing multiracial people and entrenching US-centric societal categorizations not rooted in biology (Field et al., 2021).”
The ROOTS Search Tool: Data Transparency for LLMs (2023, background): “…et al., 2021), privacy-infringing data (Carlini et al., 2020; Huang et al., 2022), incorrect language codes or and translations (Kreutzer et al., 2022), not to mention the ubiquitous issues with social biases (Blodgett et al., 2020; Field et al., 2021; Stanczak and Augenstein, 2021, among others).”
Data Governance in the Age of Large-Scale Data-Driven Language Technology (2022, background): “[53] Anjalie Field, Su Lin Blodgett, Zeerak Waseem, and Yulia Tsvetkov.”
You reap what you sow: On the Challenges of Bias Evaluation Under Multilingual Settings (2022, background): “Previous works have highlighted the importance of studying intersec-723 tional bias (Bender et al., 2021; Buolamwini and Gebru, 2018; Field et al., 2021; Guo et al., 2019; Crenshaw, 1991) but little research has been conducted around addressing such biases (Magee et al., 2021; Guo and Caliskan,…”
Gendered Mental Health Stigma in Masked Language Models (2022, background): “Most of these works study bias in a single dimension as intersectionality is difficult to operationalize (Field et al., 2021), though a”
Measuring Harmful Sentence Completion in Language Models for LGBTQIA+ Individuals (2022, background): “In this context, some dimensions of stereotyping, such as (binary) sexism (e.g., Park et al., 2018; Sun et al., 2019, inter alia ), and racism (e.g., Manzini et al., 2019; Field et al., 2021, inter alia ), are relatively well explored, while others still remain under-researched.”
Perturbation Augmentation for Fairer NLP (2022, background): “…issues can lead to unfair or poor treatment of particular demographic groups (Baro-cas et al., 2017; Mehrabi et al., 2021), a problem that is particularly egregious for historically marginalized groups, including people of color (Field et al., 2021), and women (Hendricks et al., 2018).”
Sociotechnical Harms: Scoping a Taxonomy for Harm Reduction (2022, background): “When such classi-ﬁcation relies on phenotypes, this dimension of representational harm essentializes historically contingent identities [75] through which classiﬁcation systems entrench and produce meaning about what they represent [89, 93].”
Undesirable biases in NLP: Averting a crisis of measurement (2022, background): “This is particularly problematic, considering that most research on bias in NLP is focused on one type of bias in one language: gender bias for the English language (Field et al., 2021; Talat et al., 2022).”
Biases in Large Language Models: Origins, Inventory, and Discussion (2023, background): “[51] provide a useful survey of the complexity and subtleties of race and racism in NLP.”
Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction (2022, background): “When such classification relies on phenotypes, this dimension of representational harm essentializes historically contingent identities [85] through which classification systems entrench and produce meaning about what they represent [100, 111].”
From Pretraining Data to Language Models to Downstream Tasks: Tracking the Trails of Political Biases Leading to Unfair NLP Models (2023, background): “sues in NLP models (Blodgett et al., 2020a; Field et al., 2021; Kumar et al., 2022) and designed synthetic datasets (Nangia et al.”