Word embeddings carry stereotypical connotations from the text they are trained on, which can lead to invalid inferences in downstream models that rely on them. We use this observation to design a mechanism for measuring stereotypes using the task of natural language inference. We demonstrate a reduction in invalid inferences via bias mitigation strategies on static word embeddings (GloVe). Further, we show that for gender bias, these techniques extend to contextualized embeddings when applied selectively only to the static components of contextualized embeddings (ELMo, BERT).
Word embeddings are the standard model for semantic and syntactic representations of words. Unfortunately, these models have been shown to exhibit undesirable word associations resulting from gender, racial, and religious biases. Existing post-processing methods for debiasing word embeddings are unable to mitigate gender bias hidden in the spatial arrangement of word vectors. In this paper, we propose RAN-Debias, a novel gender debiasing methodology which not only eliminates the bias present in a word vector but also alters the spatial distribution of its neighbouring vectors, achieving a bias…
Michigan State University（密歇根州立大学）；TAL Education Group（好未来教育集团）
It is evident that deep text classification models trained on human data could be biased. In particular, they produce biased outcomes for texts that explicitly include identity terms of certain demographic groups. We refer to this type of bias as explicit bias, which has been extensively studied. However, deep text classification models can also produce biased outcomes for texts written by authors of certain demographic groups. We refer to such bias as implicit bias of which we still have a rather limited understanding. In this paper, we first demonstrate that implicit bias exists in different…
Word embedding has become essential for natural language processing as it boosts empirical performances of various tasks. However, recent research discovers that gender bias is incorporated in neural word embeddings, and downstream tasks that rely on these biased word vectors also produce gender-biased results. While some word-embedding gender-debiasing methods have been developed, these methods mainly focus on reducing gender bias associated with gender direction and fail to reduce the gender bias presented in word embedding relations. In this paper, we design a causal and simple approach for…
Technion（以色列理工学院）；Israel Institute of Technology（以色列理工学院）；Harvard University（哈佛大学）
Many natural language inference (NLI) datasets contain biases that allow models to perform well by only using a biased subset of the input, without considering the remainder features. For instance, models are able to make a classification decision by only using the hypothesis, without learning the true relationship between it and the premise. These structural biases lead discriminative models to learn unintended superficial features and to generalize poorly out of the training distribution. In this work, we reformulate the NLI task as a generative task, where a model is conditioned on the bias…
University of Qom（库姆大学）；Tehran Institute for Advanced Studies, Khatam University（哈塔姆大学德黑兰高等研究院）；LMU Munich（慕尼黑大学）
Several proposals have been put forward in recent years for improving out-of-distribution (OOD) performance through mitigating dataset biases. A popular workaround is to train a robust model by re-weighting training examples based on a secondary biased model. Here, the underlying assumption is that the biased model resorts to shortcut features. Hence, those training examples that are correctly predicted by the biased model are flagged as being biased and are down-weighted during the training of the main model. However, assessing the importance of an instance merely based on the predictions of …
Most NLP datasets are not annotated with protected attributes such as gender, making it difficult to measure classification bias using standard measures of fairness (e.g., equal opportunity). However, manually annotating a large dataset with a protected attribute is slow and expensive. Instead of annotating all the examples, can we annotate a subset of them and use that sample to estimate the bias? While it is possible to do so, the smaller this annotated sample is, the less certain we are that the estimate is close to the true bias. In this work, we propose using Bernstein bounds to represent…
TU Wien（维也纳工业大学）；University College London（伦敦大学学院）
Supervised machine learning models and their evaluation strongly depends on the quality of the underlying dataset. When we search for a relevant piece of information it may appear anywhere in a given passage. However, we observe a bias in the position of the correct answer in the text in two popular Question Answering datasets used for passage re-ranking. The excessive favoring of earlier positions inside passages is an unwanted artefact. This leads to three common Transformer-based re-ranking models to ignore relevant parts in unseen passages. More concerningly, as the evaluation set is taken…
KU Leuven（鲁汶大学）；University of Antwerp（安特卫普大学）；TU Berlin（柏林工业大学）
An increasing awareness of biased patterns in natural language processing resources, like BERT, has motivated many metrics to quantify `bias' and `fairness'. But comparing the results of different metrics and the works that evaluate with such metrics remains difficult, if not outright impossible. We survey the existing literature on fairness metrics for pretrained language models and experimentally evaluate compatibility, including both biases in language models as in their downstream tasks. We do this by a mixture of traditional literature survey and correlation analysis, as well as by runnin…
LLMs Are Biased Towards Output Formats! Systematically Evaluating and Mitigating Output Format Bias of LLMs
National University of Singapore（新加坡国立大学）；VinAI Research（VinAI研究院）；Institute of High Performance Computing (IHPC), A*STAR（新加坡科技研究局高性能计算研究所）；Salesforce Research（Salesforce研究院）；Nanyang Technological University（南洋理工大学）
University of Groningen（格罗宁根大学）；University of Malta（马耳他大学）
Contextualized word embeddings have been replacing standard embeddings as the representational knowledge source of choice in NLP systems. Since a variety of biases have previously been found in standard word embeddings, it is crucial to assess biases encoded in their replacements as well. Focusing on BERT (Devlin et al., 2018), we measure gender bias by studying associations between gender-denoting target words and names of professions in English and German, comparing the findings with real-world workforce statistics. We mitigate bias by fine-tuning BERT on the GAP corpus (Webster et al., 2018…
The University of Tokyo（东京大学）；National Institute of Informatics（国立信息学研究所）
Extractive question answering (QA) models tend to exploit spurious correlations to make predictions when a training set has unintended biases. This tendency results in models not being generalizable to examples where the correlations do not hold. Determining the spurious correlations QA models can exploit is crucial in building generalizable QA models in real-world applications; moreover, a method needs to be developed that prevents these models from learning the spurious correlations even when a training set is biased. In this study, we discovered that the relative position of an answer, whic…
Many modern Artificial Intelligence (AI) systems make use of data embeddings, particularly in the domain of Natural Language Processing (NLP). These embeddings are learnt from data that has been gathered "from the wild" and have been found to contain unwanted biases. In this paper we make three contributions towards measuring, understanding and removing this problem. We present a rigorous way to measure some of these biases, based on the use of word lists created for social psychology applications; we observe how gender bias in occupations reflects actual gender bias in the same occupations in…
University of Qom（库姆大学）；Tehran Institute for Advanced Studies（德黑兰高等研究院）；Khatam University（哈塔姆大学）
Existing techniques for mitigating dataset bias often leverage a biased model to identify biased instances. The role of these biased instances is then reduced during the training of the main model to enhance its robustness to out-of-distribution data. A common core assumption of these techniques is that the main model handles biased instances similarly to the biased model, in that it will resort to biases whenever available. In this paper, we show that this assumption does not hold in general. We carry out a critical investigation on two well-known datasets in the domain, MNLI and FEVER, along…
University of Cambridge（剑桥大学）；Bar-Ilan University（巴伊兰大学）
This paper treats gender bias latent in word embeddings. Previous mitigation attempts rely on the operationalisation of gender bias as a projection over a linear subspace. An alternative approach is Counterfactual Data Augmentation (CDA), in which a corpus is duplicated and augmented to remove bias, e.g. by swapping all inherently-gendered words in the copy. We perform an empirical comparison of these approaches on the English Gigaword and Wikipedia, and find that whilst both successfully reduce direct bias and perform well in tasks which quantify embedding quality, CDA variants outperform pro…
Language representations are known to carry stereotypical biases and, as a result, lead to biased predictions in downstream tasks. While existing methods are effective at mitigating biases by linear projection, such methods are too aggressive: they not only remove bias, but also erase valuable information from word embeddings. We develop new measures for evaluating specific information retention that demonstrate the tradeoff between bias removal and information retention. To address this challenge, we propose OSCaR (Orthogonal Subspace Correction and Rectification), a bias-mitigating method th…
Counterspeech for Mitigating the Influence of Media Bias: Comparing Human and LLM-Generated Responses
The Chinese University of Hong Kong, Hong Kong, China（香港中文大学）；MoE Key Laboratory of High Confidence Software Technologies（高可信软件技术国家重点实验室）；Harbin Institute of Technology, Shenzhen, China（哈尔滨工业大学（深圳））
Measuring and Mitigating Post-hoc Rationalization in Reverse Chain-of-Thought Generation
State Key Laboratory of Multimedia Information Processing, School of Computer Science, Peking University（信息处理国家重点实验室，计算机科学学院，北京大学）；University of Electronic Science and Technology of China（电子科技大学）；Nanbeige Lab, BOSS Zhipin（纳贝格实验室，BOSS智联）
Same Question, Different Answer? Measuring and Mitigating Prompt Privilege for Equitable AI Access
Fuqua School of Business Duke University（杜克大学福库商学院）；Department of Engineering Carnegie Mellon University（卡内基梅隆大学工程学院）；Department of Industrial Engineering and Management Sciences Northwestern University（西北大学工业工程与管理科学系）；Department of Information and Decision Sciences University of Minnesota（明尼苏达大学信息与决策科学学院）
TriCon-Fair: Triplet Contrastive Learning for Mitigating Social Bias in Pre-trained Language Models
School of Computer Science（计算机科学学院）；Artificial Intelligence, Wuhan University of Technology, Wuhan, China（武汉理工大学人工智能学院）；Faculty of Data Science, City University of Macau, Macau, China（澳门城市大学数据科学学院）
Shenzhen Campus of Sun Yat-sen University（中山大学深圳校区）；Sun Yat-sen University（中山大学）；Dark Matter AI Inc.（暗物质人工智能公司）
In this paper, we revisit the solving bias when evaluating models on current Math Word Problem (MWP) benchmarks. However, current solvers exist solving bias which consists of data bias and learning bias due to biased dataset and improper training strategy. Our experiments verify MWP solvers are easy to be biased by the biased training datasets which do not cover diverse questions for each problem narrative of all MWPs, thus a solver can only learn shallow heuristics rather than deep semantics for understanding problems. Besides, an MWP can be naturally solved by multiple equivalent equations w…
University of Southern California（南加州大学）；Information Sciences Institute（信息科学研究所）；Amazon Alexa（亚马逊Alexa）；Harvard University（哈佛大学）；University of California, Los Angeles（加州大学洛杉矶分校）
Language models excel at generating coherent text, and model compression techniques such as knowledge distillation have enabled their use in resource-constrained settings. However, these models can be biased in multiple ways, including the unfounded association of male and female genders with gender-neutral professions. Therefore, knowledge distillation without any fairness constraints may preserve or exaggerate the teacher model's biases onto the distilled model. To this end, we present a novel approach to mitigate gender disparity in text generation by learning a fair model during knowledge …
Dartmouth College（达特茅斯学院）；University of Texas at Austin（德克萨斯大学奥斯汀分校）；ProtagoLabs
Current large-scale language models can be politically biased as a result of the data they are trained on, potentially causing serious problems when they are deployed in real-world settings. In this paper, we describe metrics for measuring political bias in GPT-2 generation and propose a reinforcement learning (RL) framework for mitigating political biases in generated text. By using rewards from word embeddings or a classifier, our RL framework guides debiased generation without having access to the training data or requiring the model to be retrained. In empirical experiments on three attrib…
Michigan State University（密歇根州立大学）；TAL Education Group（好未来教育集团）
Dialogue systems play an increasingly important role in various aspects of our daily life. It is evident from recent research that dialogue systems trained on human conversation data are biased. In particular, they can produce responses that reflect people's gender prejudice. Many debiasing methods have been developed for various NLP tasks, such as word embedding. However, they are not directly applicable to dialogue systems because they are likely to force dialogue models to generate similar responses for different genders. This greatly degrades the diversity of the generated responses and im…
Center for Artificial Intelligence Research (CAiRE)（人工智能研究中心（CAiRE））；Hong Kong University of Science and Technology（香港科技大学）
Media bias can lead to increased political polarization, and thus, the need for automatic mitigation methods is growing. Existing mitigation work displays articles from multiple news outlets to provide diverse news coverage, but without neutralizing the bias inherent in each of the displayed articles. Therefore, we propose a new task, a single neutralized article generation out of multiple biased articles, to facilitate more efficient access to balanced and unbiased information. In this paper, we compile a new dataset NeuWS, define an automatic evaluation metric, and provide baselines and mult…
