Recent studies show that Natural Language Processing (NLP) technologies propagate societal biases about demographic groups associated with attributes such as gender, race, and nationality. To create interventions and mitigate these biases and associated harms, it is vital to be able to detect and measure such biases. While existing works propose bias evaluation and mitigation methods for various tasks, there remains a need to cohesively understand the biases and the specific harms they measure, and how different measures compare with each other. To address this gap, this work presents a practical framework of harms and a series of questions that practitioners can answer to guide the development of bias measures. As a validation of our framework and documentation questions, we also present several case studies of how existing bias measures in NLP -- both intrinsic measures of bias in representations and extrinsic measures of bias of downstream applications -- can be aligned with different harms and how our proposed documentation questions facilitates more holistic understanding of what bias measures are measuring.
Measuring bias is key for better understanding and addressing unfairness in NLP/ML models. This is often done via fairness metrics which quantify the differences in a model's behaviour across a range of demographic groups. In this work, we shed more light on the differences and similarities between the fairness metrics used in NLP. First, we unify a broad range of existing metrics under three generalized fairness metrics, revealing the connections between them. Next, we carry out an extensive empirical comparison of existing metrics and demonstrate that the observed differences in bias measure…
Undesirable Biases in NLP: Addressing Challenges of Measurement
Institute for Logic, Language and Computation, University of Amsterdam（阿姆斯特丹大学逻辑、语言与计算研究所）；Department of Experimental Psychology, Utrecht University（乌得勒支大学实验心理学系）
University of Oxford（牛津大学）；The Alan Turing Institute（艾伦·图灵研究所）；Mozilla Foundation（Mozilla基金会）；University College Dublin（都柏林大学学院）；IT University of Copenhagen（哥本哈根信息技术大学）
Text data can pose a risk of harm. However, the risks are not fully understood, and how to handle, present, and discuss harmful text in a safe way remains an unresolved issue in the NLP community. We provide an analytical framework categorising harms on three axes: (1) the harm type (e.g., misinformation, hate speech or racial stereotypes); (2) whether a harm is \textit{sought} as a feature of the research design if explicitly studying harmful content (e.g., training a hate speech classifier), versus \textit{unsought} if harmful content is encountered when working on unrelated problems (e.g., …
School of Computing and Information Systems（计算与信息系统学院）；University of Melbourne（墨尔本大学）
Bias is pervasive in NLP models, motivating the development of automatic debiasing techniques. Evaluation of NLP debiasing methods has largely been limited to binary attributes in isolation, e.g., debiasing with respect to binary gender or race, however many corpora involve multiple such attributes, possibly with higher cardinality. In this paper we argue that a truly fair model must consider `gerrymandering' groups which comprise not only single attributes, but also intersectional groups. We evaluate a form of bias-constrained model which is new to NLP, as well an extension of the iterative n…
Perturbation Sensitivity Analysis to Detect Unintended Model Biases
Google Brain（谷歌大脑）
数据驱动的统计自然语言处理（NLP）技术利用大量语言数据来构建能够理解语言的模型。然而，大多数语言数据反映了数据产生时的公共话语，因此，除了一般的语言意义外，NLP模型还容易在特定时间点学习围绕命名对象的偶然关联。一个NLP系统设计用于对情感和毒性等概念进行建模，理想情况下，它应该产生独立于文本中提到的这些实体的身份及其社会关联的分数。例如，在通用情感分析系统中，诸如I hate Katy Perry的短语应当被解释为具有与I hate Taylor Swift相同的情感。基于这一思想，我们提出了一个通用的评估框架，扰动敏感性分析，检测与命名实体相关的非预期模型偏差，并且不需要新的注释或语料库。我们证明了这种分析的效用，采用它对两个不同的NLP模型-情绪模型和毒性模型-应用于在线评论英语语言从四个不同的流派。
Examining risks of racial biases in NLP tools for child protective services
Johns Hopkins University（约翰斯·霍普金斯大学）；Carnegie Mellon University（卡内基梅隆大学）；Microsoft Research NYC（微软纽约研究院）；The University of North Carolina at Chapel Hill（北卡罗来纳大学教堂山分校）；University of Washington（华盛顿大学）
University of Illinois Chicago（伊利诺伊大学芝加哥分校）；University of Illinois Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）；Arizona State University（亚利桑那州立大学）
Natural Language Processing (NLP) models have been found discriminative against groups of different social identities such as gender and race. With the negative consequences of these undesired biases, researchers have responded with unprecedented effort and proposed promising approaches for bias mitigation. In spite of considerable practical importance, current algorithmic fairness literature lacks an in-depth understanding of the relations between different forms of biases. Social bias is complex by nature. Numerous studies in social psychology identify the "generalized prejudice", i.e., gene…
An increasing number of works in natural language processing have addressed the effect of bias on the predicted outcomes, introducing mitigation techniques that act on different parts of the standard NLP pipeline (data and models). However, these works have been conducted in isolation, without a unifying framework to organize efforts within the field. This leads to repetitive approaches, and puts an undue focus on the effects of bias, rather than on their origins. Research focused on bias symptoms rather than the underlying origins could limit the development of effective countermeasures. In t…
We analyze 6.7 million case law documents to determine the presence of gender bias within our judicial system. We find that current bias detectino methods in NLP are insufficient to determine gender bias in our case law database and propose an alternative approach. We show that existing algorithms' inconsistent results are consequences of prior research's definition of biases themselves. Bias detection algorithms rely on groups of words to represent bias (e.g., 'salary,' 'job,' and 'boss' to represent employment as a potentially biased theme against women in text). However, the methods to buil…
University of Sheffield（谢菲尔德大学）；Humboldt University, Berlin（柏林洪堡大学）；Hero I/S；University of Copenhagen（哥本哈根大学）
Machine Learning seeks to identify and encode bodies of knowledge within provided datasets. However, data encodes subjective content, which determines the possible outcomes of the models trained on it. Because such subjectivity enables marginalisation of parts of society, it is termed (social) `bias' and sought to be removed. In this paper, we contextualise this discourse of bias in the ML community against the subjective choices in the development process. Through a consideration of how choices in data and model development construct subjectivity, or biases that are represented in a model, we…
We propose a bias-aware methodology to engage with power relations in natural language processing (NLP) research. NLP research rarely engages with bias in social contexts, limiting its ability to mitigate bias. While researchers have recommended actions, technical methods, and documentation practices, no methodology exists to integrate critical reflections on bias with technical NLP methods. In this paper, after an extensive and interdisciplinary literature review, we contribute a bias-aware methodology for NLP research. We also contribute a definition of biased text, a discussion of the impli…
Language data and models demonstrate various types of bias, be it ethnic, religious, gender, or socioeconomic. AI/NLP models, when trained on the racially biased dataset, AI/NLP models instigate poor model explainability, influence user experience during decision making and thus further magnifies societal biases, raising profound ethical implications for society. The motivation of the study is to investigate how AI systems imbibe bias from data and produce unexplainable discriminatory outcomes and influence an individual's articulateness of system outcome due to the presence of racial bias fea…
University of Alberta（阿尔伯塔大学）；University of Essex（埃塞克斯大学）；University of Texas at Arlington（德克萨斯大学阿灵顿分校）
With widening deployments of natural language processing (NLP) in daily life, inherited social biases from NLP models have become more severe and problematic. Previous studies have shown that word embeddings trained on human-generated corpora have strong gender biases that can produce discriminative results in downstream tasks. Previous debiasing methods focus mainly on modeling bias and only implicitly consider semantic information while completely overlooking the complex underlying causal structure among bias and semantic components. To address these issues, we propose a novel methodology th…
As NLP models become more integrated with the everyday lives of people, it becomes important to examine the social effect that the usage of these systems has. While these models understand language and have increased accuracy on difficult downstream tasks, there is evidence that these models amplify gender, racial and cultural stereotypes and lead to a vicious cycle in many settings. In this survey, we analyze the origins of biases, the definitions of fairness, and how different subfields of NLP mitigate bias. We finally discuss how future studies can work towards eradicating pernicious biases…
Center for Information and Language Processing (CIS), LMU Munich（慕尼黑大学信息与语言处理中心）；Institute of Social and Cultural Anthropology, LMU Munich（慕尼黑大学社会与文化人类学研究所）
When trained on large, unfiltered crawls from the internet, language models pick up and reproduce all kinds of undesirable biases that can be found in the data: they often generate racist, sexist, violent or otherwise toxic language. As large models require millions of training examples to achieve good performance, it is difficult to completely prevent them from being exposed to such content. In this paper, we first demonstrate a surprising finding: pretrained language models recognize, to a considerable degree, their undesirable biases and the toxicity of the content they produce. We refer to…
Are Non-English Papers Reviewed Fairly? Language-of-Study Bias in NLP Peer Reviews
Koç University（科奇大学）；KUIS AI Lab（KUIS人工智能实验室）；LMU Munich（慕尼黑大学）；Munich Center for Machine Learning（慕尼黑机器学习中心）；UCLouvain（鲁汶大学）；Friedrich-Alexander-Universität Erlangen-Nürnberg（埃尔朗根-纽伦堡大学）
Various existing studies have analyzed what social biases are inherited by NLP models. These biases may directly or indirectly harm people, therefore previous studies have focused only on human attributes. However, until recently no research on social biases in NLP regarding nonhumans existed. In this paper, we analyze biases to nonhuman animals, i.e. speciesist bias, inherent in English Masked Language Models such as BERT. We analyzed speciesist bias against 46 animal names using template-based and corpus-extracted sentences containing speciesist (or non-speciesist) language. We found that pr…
Motivations for methods in explainable artificial intelligence (XAI) often include detecting, quantifying and mitigating bias, and contributing to making machine learning models fairer. However, exactly how an XAI method can help in combating biases is often left unspecified. In this paper, we briefly review trends in explainability and fairness in NLP research, identify the current practices in which explainability methods are applied to detect and mitigate bias, and investigate the barriers preventing XAI methods from being used more widely in tackling fairness issues.
College of Information and Computer Sciences（信息与计算机科学学院）；University of Massachusetts Amherst（马萨诸塞大学阿默斯特分校）；Microsoft Research（微软研究院）；Cornell University（康奈尔大学）；University of Maryland（马里兰大学）
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 …
The rise of concern around Natural Language Processing (NLP) technologies containing and perpetuating social biases has led to a rich and rapidly growing area of research. Gender bias is one of the central biases being analyzed, but to date there is no comprehensive analysis of how "gender" is theorized in the field. We survey nearly 200 articles concerning gender bias in NLP to discover how the field conceptualizes gender both explicitly (e.g. through definitions of terms) and implicitly (e.g. through how gender is operationalized in practice). In order to get a better idea of emerging trajec…
Tokyo Institute of Technology（东京工业大学）；University of Liverpool（利物浦大学）；Amazon（亚马逊）
Masked Language Models (MLMs) have shown superior performances in numerous downstream NLP tasks when used as text encoders. Unfortunately, MLMs also demonstrate significantly worrying levels of social biases. We show that the previously proposed evaluation metrics for quantifying the social biases in MLMs are problematic due to following reasons: (1) prediction accuracy of the masked tokens itself tend to be low in some MLMs, which raises questions regarding the reliability of the evaluation metrics that use the (pseudo) likelihood of the predicted tokens, and (2) the correlation between the p…
University of Southern California（南加州大学）；Sony Group Corporation（索尼集团公司）
NLP models are susceptible to learning spurious biases (i.e., bugs) that work on some datasets but do not properly reflect the underlying task. Explanation-based model debugging aims to resolve spurious biases by showing human users explanations of model behavior, asking users to give feedback on the behavior, then using the feedback to update the model. While existing model debugging methods have shown promise, their prototype-level implementations provide limited practical utility. Thus, we propose XMD: the first open-source, end-to-end framework for explanation-based model debugging. Given …
The University of Melbourne（墨尔本大学）；MBZUAI（穆罕默德·本·扎耶德人工智能大学）
Modern NLP systems exhibit a range of biases, which a growing literature on model debiasing attempts to correct. However current progress is hampered by a plurality of definitions of bias, means of quantification, and oftentimes vague relation between debiasing algorithms and theoretical measures of bias. This paper seeks to clarify the current situation and plot a course for meaningful progress in fair learning, with two key contributions: (1) making clear inter-relations among the current gamut of methods, and their relation to fairness theory; and (2) addressing the practical problem of mod…
Researchers in NLP often frame and discuss research results in ways that serve to deemphasize the field's successes, often in response to the field's widespread hype. Though well-meaning, this has yielded many misleading or false claims about the limits of our best technology. This is a problem, and it may be more serious than it looks: It harms our credibility in ways that can make it harder to mitigate present-day harms, like those involving biased systems for content moderation or resume screening. It also limits our ability to prepare for the potentially enormous impacts of more distant fu…
Recent research has revealed undesirable biases in NLP data and models. However, these efforts largely focus on social disparities in the West, and are not directly portable to other geo-cultural contexts. In this position paper, we outline a holistic research agenda to re-contextualize NLP fairness research for the Indian context, accounting for Indian societal context, bridging technological gaps in capability and resources, and adapting to Indian cultural values. We also summarize findings from an empirical study on various social biases along different axes of disparities relevant to India…
University of Calicut（卡利卡特大学）；Queen’s University Belfast（贝尔法斯特女王大学）
The remarkable progress in Natural Language Processing (NLP) brought about by deep learning, particularly with the recent advent of large pre-trained neural language models, is brought into scrutiny as several studies began to discuss and report potential biases in NLP applications. Bias in NLP is found to originate from latent historical biases encoded by humans into textual data which gets perpetuated or even amplified by NLP algorithm. We present a survey to comprehend bias in large pre-trained language models, analyze the stages at which they occur in these models, and various ways in whic…
Colombian Waitresses y Jueces canadienses: Gender and Country Biases in Occupation Recommendations from LLMs
Erasmus Mundus Master in Language and Communication Technologies（埃拉斯谟 Mundus 语言与沟通技术硕士项目）；HiTZ Center - Ixa, University of the Basque Country (UPV/EHU)（巴斯克国家大学（UPV/EHU）HiTZ中心 - Ixa）；Cohere（Cohere公司）；Saarland University（萨尔兰州大学）
University of Southern California（南加州大学）；Information Sciences Institute（信息科学研究所）
Warning: this paper contains content that may be offensive or upsetting. Numerous natural language processing models have tried injecting commonsense by using the ConceptNet knowledge base to improve performance on different tasks. ConceptNet, however, is mostly crowdsourced from humans and may reflect human biases such as "lawyers are dishonest." It is important that these biases are not conflated with the notion of commonsense. We study this missing yet important problem by first defining and quantifying biases in ConceptNet as two types of representational harms: overgeneralization of polar…
A Toolbox for Surfacing Health Equity Harms and Biases in Large Language Models
Google Research（谷歌研究院）；Google DeepMind（谷歌DeepMind）；University of Alberta（阿尔伯塔大学）；Massachusetts Institute of Technology（麻省理工学院）；Beth Israel Deaconess Medical Center（贝斯以色列女执事医疗中心）；Harvard T.H. Chan School of Public Health（哈佛陈曾熙公共卫生学院）
Freie Universität Berlin（柏林自由大学）；University of Southern Denmark（南丹麦大学）；Universität der Bundeswehr München（慕尼黑联邦国防军大学）
Hate speech detection is a common downstream application of natural language processing (NLP) in the real world. In spite of the increasing accuracy, current data-driven approaches could easily learn biases from the imbalanced data distributions originating from humans. The deployment of biased models could further enhance the existing social biases. But unlike handling tabular data, defining and mitigating biases in text classifiers, which deal with unstructured data, are more challenging. A popular solution for improving machine learning fairness in NLP is to conduct the debiasing process wi…
As language models grow in popularity, it becomes increasingly important to clearly measure all possible markers of demographic identity in order to avoid perpetuating existing societal harms. Many datasets for measuring bias currently exist, but they are restricted in their coverage of demographic axes and are commonly used with preset bias tests that presuppose which types of biases models can exhibit. In this work, we present a new, more inclusive bias measurement dataset, HolisticBias, which includes nearly 600 descriptor terms across 13 different demographic axes. HolisticBias was assembl…
Several benchmarks have been built with heavy investment in resources to track our progress in NLP. Thousands of papers published in response to those benchmarks have competed to top leaderboards, with models often surpassing human performance. However, recent studies have shown that models triumph over several popular benchmarks just by overfitting on spurious biases, without truly learning the desired task. Despite this finding, benchmarking, while trying to tackle bias, still relies on workarounds, which do not fully utilize the resources invested in benchmark creation, due to the discardin…
Indian Institute of Technology Kanpur（印度理工学院坎普尔分校）；University of California, Los Angeles（加利福尼亚大学洛杉矶分校）
Language representations are efficient tools used across NLP applications, but they are strife with encoded societal biases. These biases are studied extensively, but with a primary focus on English language representations and biases common in the context of Western society. In this work, we investigate biases present in Hindi language representations with focuses on caste and religion-associated biases. We demonstrate how biases are unique to specific language representations based on the history and culture of the region they are widely spoken in, and how the same societal bias (such as bin…
We investigate the impact of political ideology biases in training data. Through a set of comparison studies, we examine the propagation of biases in several widely-used NLP models and its effect on the overall retrieval accuracy. Our work highlights the susceptibility of large, complex models to propagating the biases from human-selected input, which may lead to a deterioration of retrieval accuracy, and the importance of controlling for these biases. Finally, as a way to mitigate the bias, we propose to learn a text representation that is invariant to political ideology while still judging t…
