Bias and Fairness in Large Language Models: A Survey
Open accessThe DOI matches its Crossref record; 313 of the 313 references it cites were found in a registry or an index; 151 of the 183 sources its claims cite were checked for support.
Abstract
Rapid advancements of large language models (LLMs) have enabled the processing, understanding, and generation of human-like text, with increasing integration into systems that touch our social sphere. Despite this success, these models can learn, perpetuate, and amplify harmful social biases. In this article, we present a comprehensive survey of bias evaluation and mitigation techniques for LLMs. We first consolidate, formalize, and expand notions of social bias and fairness in natural language processing, defining distinct facets of harm and introducing several desiderata to operationalize fairness for LLMs. We then unify the literature by proposing three intuitive taxonomies, two for bias evaluation, namely, metrics and datasets, and one for mitigation. Our first taxonomy of metrics for bias evaluation disambiguates the relationship between metrics and evaluation datasets, and organizes metrics by the different levels at which they operate in a model: embeddings, probabilities, and generated text. Our second taxonomy of datasets for bias evaluation categorizes datasets by their structure as counterfactual inputs or prompts, and identifies the targeted harms and social groups; we also release a consolidation of publicly available datasets for improved access. Our third taxonomy of techniques for bias mitigation classifies methods by their intervention during pre-processing, in-training, intra-processing, and post-processing, with granular subcategories that elucidate research trends. Finally, we identify open problems and challenges for future work. Synthesizing a wide range of recent research, we aim to provide a clear guide of the existing literature that empowers researchers and practitioners to better understand and prevent the propagation of bias in LLMs.
Overview
CoChat Summary
TL;DR
This survey consolidates, formalizes, and taxonomizes bias evaluation and mitigation for large language models, covering metrics, datasets, and mitigation techniques. It proposes three taxonomies and identifies open problems around power imbalances, fairness conceptualization, evaluation standards, mitigation scalability, and theoretical guarantees.
Key findings
- The survey proposes three taxonomies: bias evaluation metrics organized by what they use from a model (embeddings, probabilities, generated text); bias evaluation datasets organized by structure (counterfactual inputs vs. prompts); and bias mitigation techniques organized by intervention stage (pre-processing, in-training, intra-processing, post-processing).
- It formalizes social bias for LLMs as representational harms (e.g., derogatory language, disparate system performance, erasure, exclusionary norms, misrepresentation, stereotyping, toxicity) and allocational harms (direct and indirect discrimination), and introduces fairness desiderata including fairness through unawareness, invariance, equal social group associations, equal neutral associations, and replicated distributions.
- It reports limitations across metric classes: embedding-based and probability-based metrics may be only weakly or inconsistently correlated with downstream task bias, and generated-text metrics can be sensitive to decoding parameters; it also highlights reliability and validity issues in widely used datasets.
- It consolidates publicly available datasets, with reported sizes including CrowS-Pairs (1,508), StereoSet (16,995), BBQ (58,492), RealToxicityPrompts (100,000), BOLD (23,679), HolisticBias (460,000), and Bias NLI (5,712,066).
Methods
- Design: a comprehensive survey and taxonomy-building effort rather than an empirical study, synthesizing prior work on bias evaluation and mitigation for LLMs.
- Scope: focuses on bias issues in LLMs for English, with additional languages for machine translation and multilingual models, and restricts inclusion to works that propose novel closed-form metrics, datasets, or mitigation techniques and specify applicability to social bias or fairness.
- Analysis: organizes metrics, datasets, and mitigation techniques into taxonomies, formalizes several metrics and mitigation objectives with unified notation, discusses limitations of each class, and outlines open problems and recommendations.
Limitations
- The paper states that it aligns with a technical solutionist perspective rather than a critical theoretical one, and that its taxonomies are organized by technical implementation details instead of downstream usage contexts or harms.
- The paper states that it focuses narrowly on model training and evaluation, rather than a more holistic pipeline including problem formulation, data collection, and deployment/integration.
- The paper states that it is limited to English language papers; evident from the text, it also does not report a systematic search protocol, search dates, or screening counts.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures










Verification
Claims the paper makes
121 Valid · 12 Not valid · 18 Not checked · 106 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Despite the growing emphasis on addressing these issues, bias and fairness research in LLMs often fails to precisely describe the harms of model behaviors: who is harmed, why the behavior is harmful, and how the harm reflects and reinforces social principles or hierarchies (Blodgett et al. 2020).”
What the sources say
- Partly supportedcites Blodgett et al. 2020 matched to “Language (Technology) is Power: A Critical Survey of “Bias” in NLP” (2020) sourceExperimental — accuracy not yet measured
Section 3.1 (Motivations; page not provided) finds that papers on bias in NLP often fail to explain which model behaviors are harmful, how, to whom, and why. Sections 4.1–4.2 call for attention to social hierarchies and explicit normative reasoning. The survey concerns NLP research broadly, however, and does not establish that this failure is common specifically in LLM research. (LLM, conf=0.86)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Similar to the shortcomings of embedding-based metrics, Delobelle et al. (2022) and Kaneko, Bollegala, and Okazaki (2022) point out that probability-based metrics may be only weakly correlated with biases that appear in downstream tasks”
- Also stated in: § 3.6 Recommendations · p. 27 · plain
“Probability-based metrics also show weak correlations with downstream biases.”
What the sources say
- Supportedcites Delobelle et al. 2022 matched to “Measuring Fairness with Biased Rulers: A Comparative Study on Bias Metrics for Pre-trained Language Models” (2022) source
Section 3 describes CrowS-Pairs as a pseudo-loglikelihood-based metric. Section 4.4 reports that CrowS-Pairs correlates less with two measures of bias in downstream tasks, supporting the hedged claim that probability-based metrics may correlate weakly with downstream bias. The finding does not establish that all probability-based metrics do so. (LLM, conf=0.90)
- Supportedcites Kaneko, Bollegala, and Okazaki 2022 matched to “Debiasing isn't enough! -- On the Effectiveness of Debiasing MLMs and their Social Biases in Downstream Tasks” (2022) source
The Introduction (p. 1) identifies intrinsic measures as using likelihood scores assigned by MLMs and reports only weak correlations with downstream, task-specific bias measures. Section 5.2 (p. 8) likewise finds weak and inconsistent correlations between intrinsic and extrinsic bias scores. (LLM, conf=0.93)
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § 3.6 Recommendations · p. 27 · plain
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Many approaches, for instance, assume some implicitly desirable criterion ( e.g. , a model output should be independent of any social group in the input), but do not explicitly acknowledge or state the normative social values that justify their framework.”
What the sources say
- Partly supportedcites Blodgett et al. 2021 matched to “Stereotyping Norwegian Salmon: An Inventory of Pitfalls in Fairness Benchmark Datasets” (2021) sourceExperimental — accuracy not yet measured
Section 2.3 describes assumptions about ideal model behavior, including equal probabilities for sentences about different target groups. Section 6 says the benchmarks make implicit decisions about stereotyping that are not explicitly discussed or justified. The paper examines four stereotyping benchmarks, however, rather than establishing this claim about many bias and fairness approaches generally or specifically about their normative social values. (LLM, conf=0.84)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Cabello, Jørgensen, and Søgaard (2023) illustrate that associations between the representations of protected attribute and other words can be independent of downstream performance disparities, if certain assumptions of social groups' language use are violated.”
What the sources say
- Supportedcites Cabello, Jørgensen, and Søgaard 2023 matched to “On the independence of association bias and empirical fairness in language models” (2023) source
The Bias section defines association bias as differences in how demographic-group terms are encoded and empirical fairness as performance differences across social groups. Section 5 says the two can be independent and explains that linking them relies on the In-Group Affinity Assumption about how groups use those terms; it calls that assumption problematic. No page numbers are provided in our copy. (LLM, conf=0.93)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Each stakeholder brings perspectives that may specify different fairness constraints for the same application and setting.”
What the sources say
- Partly supportedcites Jernite et al. (2022) matched to “Data Governance in the Age of Large-Scale Data-Driven Language Technology” (2022) sourceExperimental — accuracy not yet measured
Section 2.2 (p. 4) says participants bring different perspectives and warns that relying only on their values could exclude views more relevant to other language users. The paper discusses nondiscrimination and local expressions of values, but does not say that stakeholders specify different fairness constraints for the same application and setting. (LLM, conf=0.85)
- Not matchedcites Bender (2019)
An author–year citation the paper’s reference list doesn’t settle. We don’t search these by name.
hedgedstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
71 more claims · 34 with their cited sources checked
Create a free account to see all 257 claims and every source verdict.
- Probably not a substantial claim. 151
- It may state two separate claims as one. 70
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“While this process mitigates some risks inherent to AI-generated content, it does not eliminate potential biases associated with proprietary large language models, which lack transparency regarding training data and internal parameters [10].”
From AI-Generated Multiple Mini Interview (MMI) Stations for Medical School Admissions: Psychometric Evaluation (2026) · cites this paper as [10]
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“Notably, there is growing attention on assessing LLM bias [63-64] and robustness [323-324].”
From A Survey of Large Language Models (2026) · cites this paper as [63]
The cited source says what the claim attributes to it.
In the citing paper
“these models pose serious risks, as the vast, unfiltered internet data used to train them can reinforce and amplify harmful biases, disproportionately harming marginalized communities (73).”
From Emergent social conventions and collective bias in LLM populations (2025) · cites this paper as [73]
The cited source says what the claim attributes to it.
In the citing paper
“Despite the use of various mitigation techniques, such as debiasing algorithms, curating balanced datasets, or incorporating fairness-focused training objectives, eliminating bias entirely is a persistent challenge141–143.”
From Current applications and challenges in large language models for patient care: a systematic review (2025) · cites this paper as [142]
The cited source says what the claim attributes to it.
In the citing paper
“Another major concern with LLMs is their propensity to amplify historical biases.”
From Evaluation of Large Language Models for Peer Review in Transplantation Research: Algorithm Validation Study (2026) · cites this paper as [30]
1 more claim about this paper
Create a free account to see every claim other papers make about this one.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ 4.1 Counterfactual Inputs · p. 28
“| WinoBias+ | 1,367”
§ 4.1 Counterfactual Inputs · p. 29
“WinoBias+ (Vanmassenhove, Emmery, and Shterionov 2021) augments WinoBias with gender-neutral alternatives, similar to Winogender's neutral option, with 3,167 total instances.”
Citations
Cites
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2019) · Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) · doi:10.18653/v1/n19-1423 · Cited by 54,325 · cited 6×
“, 2020), and GPT-4 (OpenAI, 2023); autoencoding models include BERT (Devlin et al., 2019), RoBERTa (Liu et al.”
BackgroundInfluential(2014) · Cited by 33,419
A Unified Approach to Interpreting Model Predictions
Lundberg, S., & Lee, S.-I. (2017) · arXiv · doi:10.48550/arxiv.1705.07874 · Cited by 23,707
Language Models are Few-Shot Learners
Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., et al. (2020) · arXiv · doi:10.48550/arxiv.2005.14165 · Cited by 23,372 · cited 2×
“, 2019), GPT-3 (Brown et al., 2020), and GPT-4 (OpenAI, 2023); autoencoding models include BERT (Devlin et al.”
Background"Why Should I Trust You?": Explaining the Predictions of Any Classifier
Ribeiro, M. T., Singh, S., & Guestrin, C. (2016) · Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining · doi:10.1145/2939672.2939778 · Cited by 17,555
RoBERTa: A Robustly Optimized BERT Pretraining Approach
Liu, Y., Ott, M., Goyal, N., Du, J., Joshi, M., Chen, D., Levy, O., Lewis, M., Zettlemoyer, L., & Stoyanov, V. (2019) · arXiv · doi:10.48550/arxiv.1907.11692 · Cited by 16,515
Kaugeranna, R., Kaugeranna, E., & Aion, (. S. 4. 6. (2023) · DROPS (Schloss Dagstuhl – Leibniz Center for Informatics) · doi:10.4230/lipics.giscience.2023.43 · Cited by 13,690
Gauriat, C.-M., Pencolé, Y., Ribot, P., & Brouillet, G. (2024) · HAL (Le Centre pour la Communication Scientifique Directe) · doi:10.4230/oasics.dx.2024.27 · Cited by 12,586
Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
Raffel, C., Shazeer, N., Roberts, A., Lee, K., Narang, S., Matena, M., Zhou, Y., Li, W., & Liu, P. J. (2019) · arXiv · doi:10.48550/arxiv.1910.10683 · Cited by 10,929 · cited 1×
“The rise and rapid advancement of large language models (LLMs) has fundamentally changed language technologies (e.g., Brown et al., 2020; Conneau et al., 2020; Devlin et al., 2019; Lewis et al., 2020; Liu et al., 2019; OpenAI, 2023; Radford et al., 2018; 2019; Raffel et al., 2020).”
BackgroundMeasuring individual differences in implicit cognition: the implicit association test
Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998) · Journal of Personality and Social Psychology · doi:10.1037/0022-3514.74.6.1464 · Cited by 9,210 · cited 2×
“, family and occupation words), emulating the Implicit Association Test (Greenwald et al., 1998).”
Method
Research works printed but not linked: 5
- Barocas, 2019, Fairness and Machine Learning: Limitations and Opportunities
- Blodgett, Su Lin . 2021. Sociolinguistically Driven Approaches for Just Natural Language Processing. Ph.D. thesis. University of Massachusetts Amherst.
- Green, 2019, ”Good” isn’t good enough, 1
- Radford, 2019, Language models are unsupervised multitask learners, OpenAI Blog, 1, 9
- Wang, 2019, BERT has a mouth, and it must speak: BERT as a Markov random field language model, 30, 10.18653/v1/w19-2304 · doi:10.18653/v1/w19-2304
Cited by
Items 201–300 of 763
Modular Chain-of-Thought (CoT) for LLM-Based Conceptual Construction Cost Estimation
Ghimire, P., Kim, K., Stentz, T. L., & Roy, T. (2026) · Buildings · doi:10.3390/buildings16020396 · Cited by 5
When AI reviews science: Can we trust the referee?
Wang, J., Liu, Y., Xu, H., Hu, K., Di, S., Ni, W., Yue, L., Zhang, M.-L., Ren, K., & Chen, L. (2026) · The Innovation Informatics · doi:10.59717/j.xinn-inform.2026.100030 · Cited by 9
Rapp, A., Lodovico, C. D., & Di, L. (2026) · International Journal of Human-Computer Interaction · doi:10.1080/10447318.2026.2632156 · Cited by 7
“…model behavior by framing interactions as benevolent, while diverting attention from how LLMs reproduce historically situated biases and concentrate epistemic and decision-making power in the hands of a small set of actors (Brandtzaeg et al., 2025; Cuadra et al., 2024; Gallegos et al., 2024).”
A Survey of Personalization: From RAG to Agent
Li, X., Jia, P., Xu, D., Wen, Y., Zhang, Y., Zhang, W., Wang, W., Wang, Y., Du, Z., Li, X., et al. (2026) · ACM Transactions on Information Systems · doi:10.1145/3802586 · Cited by 63
Generative AI and multimodal pedagogy: implications for teaching, learning, and assessment
Istrate, O. (2026) · Frontiers in Education · doi:10.3389/feduc.2026.1828953 · Cited by 5
A semantic embedding space based on large language models for modelling human beliefs
Lee, B., Aiyappa, R., Ahn, Y., Kwak, H., & An, J. (2025) · Nature Human Behaviour · doi:10.1038/s41562-025-02228-z · Cited by 21
“Fourth, there is a concern regarding the inherent biases present in the pre-trained LLMs used in our study 44 .”
Jailbreaking large language models: models, origins, and evolution of attacks and defenses
Li, X. H., Wu, J., Zheng, Q., Wang, H., Fan, M. Y., Hu, S., Guo, J., & Ting, L. I. U. (2025) · Scientia Sinica Informationis · doi:10.1360/ssi-2024-0196 · Cited by 4
Enhancements for Developing a Comprehensive AI Fairness Assessment Standard
Agarwal, A., Kumar, M., & Nene, M. J. (2025) · International Conference on Communication Systems & Networks (COMSNETS) · doi:10.1109/comsnets63942.2025.10885551 · Cited by 11
“This method helps uncover societal biases embedded in word vectors, enabling refinement to reduce unfair associations [12].”
InfluentialActions Speak Louder than Words: Agent Decisions Reveal Implicit Biases in Language Models
Li, Y., Shirado, H., & Das, S. (2025) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3715275.3732212 · Cited by 47
Artificial Intelligence to Support Human-Provided Mental Health Treatment
Soma, C. S., Kuo, P. B., Mehta, M., Srikumar, V., Imel, Z. E., & Atkins, D. C. (2026) · Annual Review of Clinical Psychology · doi:10.1146/annurev-clinpsy-061724-075336 · Cited by 9
Evaluation of LLMs for mathematical problem solving
Wang, R., Wang, R., Shen, Y., Wu, C., Zhou, Q., & Chandra, R. (2026) · Next research. · doi:10.1016/j.nexres.2026.101705 · Cited by 22
Karakikes, A., Alexiadis, P., & Kotis, K. (2024) · SN Computer Science · doi:10.1007/s42979-024-02935-w · Cited by 6
“Significant researches indicate that LLMs can be trained on, perpetuate, and amplify social biases, while newly developed bias evaluation and mitigation techniques for LLMs are presented [81].”
Cantini, R., Cosenza, G., Orsino, A., & Talia, D. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-78977-9_4 · Cited by 23
Fede, G. D., Alrabie, L., & Andolina, S. (2025) · Springer eBooks · doi:10.1007/978-981-97-8440-0_49-1 · Cited by 5
AI biases as asymmetries: a review to guide practice
Waters, G., & Honenberger, P. (2025) · Frontiers in Big Data · doi:10.3389/fdata.2025.1532397 · Cited by 8
Standardizing Intelligence: Aligning Generative AI for Regulatory and Operational CompliancePreprint
Imperial, J. M., Jones, M. D., & Madabushi, H. T. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5146161 · Cited by 4
Large Language Model Reasoning Failures
Song, P., Han, P., & Goodman, N. (2026) · arXiv · doi:10.48550/arxiv.2602.06176 · Cited by 3
Salvi, R. C., & Bosch, N. (2025) · ACM Journal on Responsible Computing · doi:10.1145/3737882 · Cited by 4
Kharitonova, K., Pérez‐Fernández, D., Gutiérrez-Hernando, J., Gutiérrez-Fandiño, A., Callejas, Z., & Griol, D. (2025) · Future Internet · doi:10.3390/fi17080340 · Cited by 4
Wang, B., Wang, C., Zhou, Z., Sun, Y., Yang, S., Huang, Y., Cui, Y., Wang, Y., Al‐Otaibi, Y. D., Bashir, A. K., et al. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3592453 · Cited by 4
“For instance, Gallegos et al. [10] offered a comprehensive survey on bias evaluation and mitigation techniques, while Manerba et al. [20] proposed a framework that employs a perplexity-based fairness score and benchmark datasets to probe societal biases.”
Advances in bilingualism as a dynamic process: 30 years of exploration in bilingual mind and brain
Gu, C., Peng, Y., & Li, P. (2025) · Journal of Neurolinguistics · doi:10.1016/j.jneuroling.2025.101288 · Cited by 6
Audit-style framework for evaluating bias in large language models
Baldwin, P. (2025) · Frontiers in Education · doi:10.3389/feduc.2025.1592037 · Cited by 8
A systematic review of bias detection methods for non-English word embeddings and language models
Puttick, A. R., Ikae, C., Rigotti, C., Fosch‐Villaronga, E., Kharas, M. W., Søraa, R. A., & Kurpicz-Briki, M. (2025) · Artificial Intelligence Review · doi:10.1007/s10462-025-11375-8 · Cited by 10
What is a fact in the age of generative AI? Fact-checking as an epistemological lens
Dierickx, L., Opdahl, A. L., Bjerknes, F., & Lindén, C. (2026) · Information Communication & Society · doi:10.1080/1369118x.2026.2630697 · Cited by 10
“Biases may be representational, technical, or social in nature (e.g., Dwivedi et al., 2023; Ferrara, 2024; Gallegos et al., 2024; Ruggieri et al., 2024), and cannot be captured solely through output evaluation.”
Sah, C. K. (2025) · IEEE International Conference on Computer Vision Workshops · doi:10.1109/iccvw69036.2025.00289 · Cited by 5
“On the prompt engineering front, several studies explore how prompt sensitivity impacts LLM-based RecSys outcomes [7, 15, 16, 21, 48].”
Bias Patterns in the Application of LLMs for Clinical Decision Support
Poulain, R., Adiba, F. I., Fayyaz, H., & Beheshti, R. (2026) · Delaware Journal of Public Health · doi:10.32481/djph.2026.03.10 · Cited by 76
Generative AI as an Interface Between Farmers and AI Algorithms: Opportunities and Challenges
Zhang, W., & Yadav, A. (2025) · Journal of the Indian Institute of Science · doi:10.1007/s41745-025-00490-8 · Cited by 7
Dennison, D. V., Ahtisham, B., Chourasia, K., Arora, N., Singh, R., Kizilcec, R. F., Nambi, A., Ganu, T., & Vashistha, A. (2026) · Proceedings of the ACM on Human-Computer Interaction · doi:10.1145/3788074 · Cited by 37
Khapre, S., Mersha, M. A., Olalekan, Z., Shakil, H., Iskandar, S., Moustafa, A., & Kalita, J. K. (2026) · Neurocomputing · doi:10.1016/j.neucom.2026.134049 · Cited by 11
Carl, T. B., Hoff, V. V. ‘., Haas, J., Cabral, P., & Akinyemi, F. O. (2026) · Ecosystem Services · doi:10.1016/j.ecoser.2026.101871 · Cited by 3
Estrada-Araoz, E. G. (2026) · Discover Public Health · doi:10.1186/s12982-026-02266-6 · Cited by 4
“Several studies have shown that language models can reflect social biases embedded in the datasets on which they were trained, raising questions about the neutrality of their outputs [30, 31].”
Haider, B. A., Gorti, A., Chadha, A. R., & Gaur, M. (2026) · Communications in computer and information science · doi:10.1007/978-3-032-20447-9_5 · Cited by 8
Júnior, E. S. D. S. (2026) · LUMEN ET VIRTUS · doi:10.56238/levv17n63-012 · Cited by 5
Moon, J. T., Lima, N. J., Froula, E., Li, H., Newsome, J. M., Trivedi, H., Bercu, Z. L., & Gichoya, J. W. (2024) · BMJ Health & Care Informatics · doi:10.1136/bmjhci-2023-100952 · Cited by 6
Designing AI to Think with Us, Not for Us
Hutka, S. A. (2024) · Ethnographic Praxis in Industry Conference Proceedings · doi:10.1111/epic.12198 · Cited by 6
Son, N. K., Sangaiah, A. K., Paramarthika, L. K. M., Rajendran, V., Bian, G., & Alenazi, M. J. F. (2025) · IEEE Transactions on Computational Social Systems · doi:10.1109/tcss.2025.3572596 · Cited by 4
“It is also critical to acknowledge that LLMs are not neutral processors; they inherit and can amplify societal biases from their training data [80].”
Impacts of Artificial Intelligence Development on Humanity and Social Values
鍾, 俊., Tan, Y. K., & Zhou, X. (2025) · Information · doi:10.3390/info16090810 · Cited by 4
Sanders, J., Mobley, J., Miller, I., Sochacka, N. W., Jensen, P. A., & Jensen, K. (2025) · Journal of Engineering Education · doi:10.1002/jee.70037 · Cited by 5
“The data corpuses on which LLMs are constructed have been noted to possess their own social biases (Gallegos et al., 2024; Kadan et al., 2022).”
Whose Side are You on: Investigating Political Bias of Large Language Models
Pit, P., Ma, X., Conway, M., Chen, Q., Bailey, J., Pit, P., Keo, P., Diep, W., & Jiang, Y. (2025) · Lecture notes in computer science · doi:10.1007/978-981-95-4969-6_22 · Cited by 33
Utilizing Generative Artificial Intelligence in the Online Counselor Education Classroom
Brue, J. S., & Brue, J. (2024) · Journal of Technology in Counselor Education and Supervision · doi:10.61888/2692-4129.1100 · Cited by 8
Empirical Study and Mitigation Methods of Bias in LLM-Based Robots
Ren, Z. (2024) · Academic Journal of Science and Technology · doi:10.54097/re9qp070 · Cited by 3
Liu, G., Bono, C. A., & Pierri, F. (2025) · PeerJ Computer Science · doi:10.7717/peerj-cs.2694 · Cited by 14
“In the following, we focus on contributions more directly related to our work and refer readers to Gallegos et al. (2024) for a more comprehensive review of the literature on this topic.”
Scaria, N., Kennedy, S. J. J., Latinovich, T., & Subramani, D. (2025) · Computers and Education Artificial Intelligence · doi:10.1016/j.caeai.2025.100539 · Cited by 8
Narratives of Divide: The Polarizing Power of Large Language Models in a Turbulent WorldPreprint
Saqr, K. M. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5124882 · Cited by 3
LLM-assisted genre analysis: The effect of sampling temperature on reliability
Grullon-Polanco, L. (2026) · Research Methods in Applied Linguistics · doi:10.1016/j.rmal.2026.100314 · Cited by 3
A Study of Foundation Models for Large-scale Time-series Forecasting
Chen, H., Rossi, R. A., Kim, S., Mahadik, K., & Eldardiry, H. (2024) · IEEE International Conference on Big Data (BigData) · doi:10.1109/bigdata62323.2024.10825153 · Cited by 2
Self-citationExploring Occupational Gender Bias in Large Language Models of High-Resource Language
Jia, K., Liu, Z., Zhang, D., Chen, H., Luo, F., & Qi, J. (2025) · Communications in computer and information science · doi:10.1007/978-981-96-4506-0_16 · Cited by 2
Prompting Fairness: How End Users Can Mitigate Bias in AI Systems
Marsden, N. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-93415-5_4 · Cited by 3
A Platform for AI-Assisted Archival Metadata Generation
Rim, K., King, O. C., Lynch, K., Verhagen, M., & Pustejovsky, J. D. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-93160-4_12 · Cited by 2
A Comprehensive Survey on Bias and Fairness in Large Language Models
Haque, F., Xu, D., & Niu, X. (2025) · Lecture notes in computer science · doi:10.1007/978-981-96-8197-6_7 · Cited by 12
Gender Trouble in Language Models: An Empirical Audit Guided by Gender Performativity Theory
Hafner, F. S., Valdivia, A., & Rocher, L. (2025) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3715275.3732112 · Cited by 11
Elliott, J., Friese, C., Harris, G., Mann, E., & Savage, M. (2025) · Sociology · doi:10.1177/00380385251344472 · Cited by 4
“This is ethically important, as sociological research would assuredly be based on interviews with people rather than composites of online personas that are biased towards white, English-speaking people who engage in online activity that results in digital data (Gallegos et al., 2024).”
Surfacing citizens’ policy perspectives at scale in the age of large language models
Lopez-Lopez, E., & Herzog, S. M. (2025) · Behavioral Science & Policy · doi:10.1177/23794607251346311 · Cited by 4
Improving Open-Response Assessment with LearnLM
Thomas, D. R., Borchers, C., Bhushan, S., Kakarla, S., Houk, A., Abboud, R., Gupta, S., Gatz, E., & Koedinger, K. R. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-98462-4_28 · Cited by 3
Investigating Social Biases in Multimodal LLMs
Perera, M. V., Narayan, K., & Patel, V. M. (2025) · International Conference on Automatic Face and Gesture Recognition · doi:10.1109/fg61629.2025.11099161 · Cited by 3
Hogan, N. R., & Corăbian, G. (2025) · Behavioral Sciences & the Law · doi:10.1002/bsl.70014 · Cited by 6
“…ethical threats, should consider recent research indicating that output from AI models can demonstrate bias against particular social groups, in part because the “training or inference procedure itself may amplify bias, beyond what is present in the training data” (p. 1107, Gallegos et al. 2024).”
Napiwotzki, I., Laue, J., Caldarone, F., Idahl, M., Hadler, U., Amrani, H., Hildt, E., Kahl, K. G., & Nejdl, W. (2025) · JMIR Formative Research · doi:10.2196/78138 · Cited by 4
“This spontaneous adoption, combined with LLMs’ known risks and susceptibility to errors, underscores the critical and urgent need for rigorous evaluation to ensure their safe and effective application in therapeutic dialogue [9,10].”
Towards Trustworthy and Explainable-by-Design Large Language Models for Automated Teacher Assessment
Li, Y., Yang, H., & Fang, Q. (2025) · Information · doi:10.3390/info16100882 · Cited by 2
The Multifaced Nature of Bias in AI: Impact on Model Generalization, Robustness, and Fairness
Ntoutsi, E. (2025) · Springer eBooks · doi:10.1007/978-3-658-48522-1_11 · Cited by 2
El impacto de la IA generativa en la labor docente de estudiantes de bachillerato
Solis-Aviles, E. J., Bravo-Briones, D. B., Solis-Aviles, J. Y., Medina, E. J. M., & Moya-Muñoz, L. D. (2025) · Revista Científica Ciencia y Método · doi:10.55813/gaea/rcym/v3/n4/108 · Cited by 5
Detection, Classification, and Mitigation of Gender Bias in Large Language Models
Cheng, X., Zan, H., Kong, L., Song, J., & Peng, M. (2025) · Lecture notes in computer science · doi:10.1007/978-981-95-3352-7_42 · Cited by 3
Dialogues of Sense and Algorithm: Reconfiguring Arts-Based Research in the AI Era
Guo, J., & Fang, H. (2025) · Qualitative Inquiry · doi:10.1177/10778004251393201 · Cited by 3
Uddin, S., Lu, H., & Hajati, F. (2025) · AI and Ethics · doi:10.1007/s43681-025-00874-7 · Cited by 5
Diagnostic performance of four AI tools in pharmacology MCQs: Accuracy, sensitivity, and specificity
Al-Rahahleh, A. J., Rizik, M., Al-Ashwal, F. Y., & Abu-Farha, R. K. (2025) · PLoS ONE · doi:10.1371/journal.pone.0337688 · Cited by 2
“Real-time retrieval may reduce bias propagation by incorporating current, externally verified content, unlike static models that tend to reinforce training-set biases [21].”
Ideology-Based LLMs for Content Moderation
Civelli, S., Bernardelle, P., Pratama, N., & Demartini, G. (2026) · ACM Transactions on Intelligent Systems and Technology · doi:10.1145/3810946 · Cited by 5
A Cognitive Architecture for Embodied AI Based on LLM Common-Sense Knowledge
Saladino, A., Brienza, M., Suriani, V., Bloisi, D. D., & Iocchi, L. (2025) · Lecture notes in computer science · doi:10.1007/978-981-95-2398-6_23 · Cited by 2
Zhou, M., Gui, F., Ai, C., Ling, L., Zhang, X., Lu, Y., Han, D., Zhao, B., Zhong, F., Liu, J., et al. (2026) · Graefe s Archive for Clinical and Experimental Ophthalmology · doi:10.1007/s00417-025-07092-1 · Cited by 3
Envisioning generative artificial intelligence in cartography: mapmaking, map use, and ethics
Kang, Y., & Wang, C. (2026) · International Journal of Cartography · doi:10.1080/23729333.2025.2582231 · Cited by 8
“…and hallucination, which produce fabricated content without any basis in real data, bias in GenAI refers to the systematic deviation of outputs caused by skewed training data, model design, or human influence during training or prompting (Gallegos et al., 2024; C. Wang, Tang, et al., 2025).”
Towards bias-free recruitment: Adversarial contrastive tuning for LLM-based resume screening
Wu, F., Ye, A., Cai, Z., & Chen, J. (2026) · Information and Software Technology · doi:10.1016/j.infsof.2026.108040 · Cited by 7
Quevedo, D. G., & Kuri, J. (2026) · Springer eBooks · doi:10.1007/978-3-032-08872-7_2 · Cited by 4
Who Invests, Who Gets Funded: Gender and Racial Bias in LLM-Generated Investment Advice
Wang, Y. E., & Gu, K. (2026) · Journal of Business Ethics · doi:10.1007/s10551-026-06251-6 · Cited by 6
“To address this, researchers have proposed various frameworks and benchmarks to detect, evaluate, and mitigate bias in LLM behavior (Chu et al., 2024; Gallegos et al., 2024; Li et al., 2023).”
Navigating artificial intelligence narratives in Chinese media: a mixed-methods discourse analysis
Li, Y., Yu, D., & Yu, L. (2026) · Digital Scholarship in the Humanities · doi:10.1093/llc/fqag004 · Cited by 4
Too human to model: the uncanny valley of large language models in simulating human systems
Zeng, Y., Brown, C., & Rounsevell, M. (2026) · npj Complexity · doi:10.1038/s44260-026-00075-1 · Cited by 13
“However, LLMs have many widely known technical limitations, such as hallucination 12 – 14 , bias 15 – 18 , and security issues 19,20 , which may affect their integration within social simulation.”
Shehmir, S., & Kashef, R. F. (2026) · Frontiers in Computer Science · doi:10.3389/fcomp.2026.1735253 · Cited by 2
Zabihi, P., Nawara, D., Ibrahim, A., & Kashef, R. F. (2026) · Applied Sciences · doi:10.3390/app16073410 · Cited by 5
Bringing Reasoning to Generative Recommendation Through the Lens of Cascaded Ranking
Lin, X., Liu, P., Wang, W., Hu, Y., Xu, C., Feng, F., Wang, Q., & Chua, T. (2026) · International World Wide Web Conference (WWW) · doi:10.1145/3774904.3792980 · Cited by 3
Pitfalls and risks of generative AI in machine learning
Lones, M. A. (2026) · Patterns · doi:10.1016/j.patter.2026.101534 · Cited by 5
“74 AI model developers put considerable effort into sanitizing the behavior of their models, with the aim of making them work in a way that matches user expecta-tions—a process also known as alignment—but this offers no guarantees and can even introduce new biases (e.g., lexical overrepresentation;…”
A Probabilistic Greedy Attempt to Be Fair in Neural Team Recommendation
Loghmani, H. G., Saeedi, M., Rueda, G., Paul, E., & Fani, H. (2026) · Computational Intelligence · doi:10.1111/coin.70221 · Cited by 2
“̂ 𝑑 ∈ for members of protected groups to be oblivious to the protected attribute [ 41, 42, 94, 107].”
InfluentialMolahasani, M., Motamedi, A., Greenspan, M., Kim, I.-M., & Etemad, A. (2025) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51701.2025.00072 · Cited by 3
“Bias identification in VLMs. Recent research has placed considerable emphasis on identifying and assessing bias in foundation models [9], including VLMs [2, 25].”
Identifying implicit bias in LLM-based chat AI toward people with intellectual disabilities
Coffey, K. V., Krahn, G. L., Hanley, J. P., & Neely, J. E. (2026) · Disability and health journal · doi:10.1016/j.dhjo.2026.102086 · Cited by 6
“Implicit social biases are manifest as exclusionary norms, misrepresentation, and stereotyping within LLM outputs [15].”
Your LLM Has a Passport: Investigating Brand Origin Geographic Bias in E-Commerce Rankings
Betello, F., Missault, P., Czinczoll, T., & Beliuha, M. (2026) · Conference on Fairness, Accountability and Transparency · doi:10.1145/3805689.3806438 · Cited by 2
Akhund, S. A., Qazi, S., Mazhar, M. A., Shaikh, A. A., Atif, E., Shaikh, A., Shaibah, H., Musa, M. K., N.Qazi, S., Khan, M. A., et al. (2026) · Medical Education Online · doi:10.1080/10872981.2026.2684837 · Cited by 3
“Student acceptance of AI involvement in grading varies in ways that have implications for institutional trust [10], and adoption in high-stakes contexts requires evidence that LLMs perform at a level comparable to trained human examiners without introducing systematic bias [11,12].”
Local Explanations for Large Language Models: a Brief Review of Methods
Volkov, E. N., & Averkin, A. N. (2024) · IEEE International Conference on Soft Computing and Measurements (SCM) · doi:10.1109/scm62608.2024.10554222 · Cited by 3
“Consequently, models may exhibit bias regarding race, gender, age, profession, and other characteristics [4].”
BackgroundXi, M., Perera, M., Matthews, B. J., Wang, R., Weiley, V., Somarathna, R., Maqbool, H., Chen, J., Engelke, U., Anderson, S., et al. (2024) · IEEE International Symposium on Mixed and Augmented Reality Adjunct (ISMAR-Adjunct) · doi:10.1109/ismar-adjunct64951.2024.00062 · Cited by 3
“While they offer significant advancements, they raise ethical concerns related to bias, fairness, and accountability [10, 29].”
Dynamic Region-Aware Fine-Tuning: Enhancing Geographic Inclusivity in Large Language Models
Kumari, K. S., Shree, S. U., & Calarany, C. (2024) · 2024 IEEE 16th International Conference on Computational Intelligence and Communication Networks (CICN) · doi:10.1109/cicn63059.2024.10847353 · Cited by 2
Teaching AI to handle exceptions: Supervised fine-tuning with human-aligned judgment
DiSorbo, M. D., Ju, H., & Aral, S. (2026) · PNAS Nexus · doi:10.1093/pnasnexus/pgag144 · Cited by 6
“Central to agents ’ ability to act autonomously is their decision-making — the process by which they choose actions on behalf of users.”
Deep Learning Approach for Evaluating Fairness in LLMs
Kumari, L., Serasiya, S., & Bharti, S. K. (2025) · International Conference on Sustainable Energy Technologies and Computational Intelligence (SETCOM) · doi:10.1109/setcom64758.2025.10932507 · Cited by 3
“…have been done on this topic, they have covered more broad topics like deep learning, artificial intelligence or machine learning, they have evaluated bias for specific scenarios and application domains or they have taken LLMs as a whole without taking into account the sizes [9,10,11,12,13,14,15].”
Tiwari, S. M. (2025) · Lecture notes on data engineering and communications technologies · doi:10.1007/978-3-031-82062-5_10 · Cited by 3
Responsible generative AI for telecom
Chergui, H. (2025) · Elsevier eBooks · doi:10.1016/b978-0-44-329135-7.00017-5 · Cited by 2
MDAT: Model and Dictionary Augmented Translation from Indic Language to English using LLM
Kashyap, K., & Sarma, S. K. (2025) · Procedia Computer Science · doi:10.1016/j.procs.2025.04.641 · Cited by 2
Unveiling gender bias in LLM-generated hero and heroine narratives
van Blerck, I. C. E., de Lima, E. S., Neggers, M. M. E., & Calders, T. (2025) · Entertainment Computing · doi:10.1016/j.entcom.2025.100972 · Cited by 8
Reversing the alternate uses task
Scherbakova, A., Dumas, D., Kagan, S., Vlaamster, T., Acar, S., & Organisciak, P. (2025) · Thinking Skills and Creativity · doi:10.1016/j.tsc.2025.101915 · Cited by 2
Automated Bias Assessment in AI-Generated Educational Content Using CEAT Framework
Peng, J., Shen, W., Rao, J., & Lin, J. (2025) · Communications in computer and information science · doi:10.1007/978-3-031-99264-3_44 · Cited by 8
Struniawski, K., Kozera, R., & Konopka, A. (2025) · Applied Sciences · doi:10.3390/app15168860 · Cited by 4
Optimizing RLHF Reward Models with Fairness Constraints
Zhao, J., Lei, Y., Zheng, W., Lai, S., Ding, H., & Zhao, C. (2025) · International Conference on Computer Science and Blockchain (CCSB) · doi:10.1109/ccsb66722.2025.11154240 · Cited by 3
Mehrotra, S., Raghavendra, P., Prajesh, B., Kambale, H., & Majumdar, P. (2025) · ACM International Conference on AI in Finance (ICAIF) · doi:10.1145/3768292.3770355 · Cited by 7
Who's Asking? Investigating Bias Through the Lens of Disability-Framed Queries in LLMs
Hari, V., Panda, K., Panda, S., Agarwal, A., & Patel, H. L. (2025) · 2025 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW) · doi:10.1109/iccvw69036.2025.00692 · Cited by 2
“Prior research has shown that LLMs often absorb and reproduce harmful societal biases - especially those tied to gender, race, religion, and health - across both contextual embeddings and generative outputs [1, 9, 18, 22, 28].”
Addressing Gender Bias in Generative Large Language ModelsPreprint
Zhou, H., Inkpen, D. Z., & Kantarcı, B. (2024) · Research Square · doi:10.21203/rs.3.rs-4670889/v1 · Cited by 2
Regulating MultifunctionalityPreprint
Coglianese, C., & Crum, C. R. (2024) · SSRN Electronic Journal · doi:10.2139/ssrn.5059426 · Cited by 2
Beyond Words: Extending a Pragmatic View of Language to Large Language Models for Informed Consent
Sobieska, A., & Starke, G. (2025) · The American Journal of Bioethics · doi:10.1080/15265161.2025.2470669 · Cited by 2
Metadata
- Authors
- Gallegos, I. O., Rossi, R. A., Barrow, J., Tanjim, M. M., Kim, S., Dernoncourt, F., Yu, T., Zhang, R., & Ahmed, N. K.
- Affiliations
- Stanford University (US)
- Adobe Research. ryrossi@adobe.com
- Pattern Data. joe.barrow@patterndataworks.com
- Adobe Research. tanjim@adobe.com
- Adobe Research. sukim@adobe.com
- Adobe Research. dernonco@adobe.com
- Adobe Research. tyu@adobe.com
- Adobe Research. ruizhang@adobe.com
- Intel (United States) (US)
- ORCID
- Venue
- Computational Linguistics
- Publisher
- MIT Press
- Published
- 2024
- Volume
- 50
- Issue
- 3
- Pages
- 1097-1179
- ISSN
- 0891-2017 (ISSN-L)
- 1530-9312
- OpenAlex
- W4399528455
- Semantic Scholar
- bcfa73aedf1b2d1ee4f168e21298a37ac55a37f7
- Cited by
- 1,245
- References
- 280
- Citations by year
- 2026: 596 · 2025: 451 · 2024: 188
- FWCI
- 132.76
- Type
- journal article
- Language
- en
- Open access
- diamond
- doi.org
- Licence
- Locations
- Computational Linguistics (publishedVersion, cc-by-nc-nd)
- arXiv (Cornell University) (submittedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- arXiv (Cornell University)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
- Physical Sciences › Computer Science › Artificial Intelligence › Hate Speech and Cyberbullying Detection
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Keywords
- large language models, social bias, fairness, bias assessment, bias mitigation, bias metrics, data preprocessing, post-processing
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-11-02, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-11-02 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.