On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜
Open accessThe DOI matches its Crossref record; 156 of the 159 references it cites were found in a registry or an index, and 3 are not checked yet.
Abstract
The past 3 years of work in NLP have been characterized by the development and deployment of ever larger language models, especially for English. BERT, its variants, GPT-2/3, and others, most recently Switch-C, have pushed the boundaries of the possible both through architectural innovations and through sheer size. Using these pretrained models and the methodology of fine-tuning them for specific tasks, researchers have extended the state of the art on a wide array of tasks as measured by leaderboards on specific benchmarks for English. In this paper, we take a step back and ask: How big is too big? What are the possible risks associated with this technology and what paths are available for mitigating those risks? We provide recommendations including weighing the environmental and financial costs first, investing resources into curating and carefully documenting datasets rather than ingesting everything on the web, carrying out pre-development exercises evaluating how the planned approach fits into research and development goals and supports stakeholder values, and encouraging research directions beyond ever larger language models.
Checked 2026-11-22 · 3 references not checked yet
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The text we found for this paper looks like another document, so the self-check wasn’t run.
Citations
Cites
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017) · arXiv · doi:10.48550/arxiv.1706.03762 · Cited by 120,404
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 5×
“[39] in particular noted that training on a large dataset and fine-tuning for specific tasks leads to strictly increasing results on the GLUE tasks [138] for English as the hyperparameters of the model were increased.”
BackgroundMethodResultInfluentialGlove: Global Vectors for Word Representation
Pennington, J., Socher, R., & Manning, C. (2014) · Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.3115/v1/d14-1162 · Cited by 34,071
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 4×
“documents linked to from Reddit [25], plus Wikipedia and a collection”
BackgroundMethodInfluentialRoBERTa: 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
Distilling the Knowledge in a Neural Network
Hinton, G., Vinyals, O., & Dean, J. (2015) · arXiv · doi:10.48550/arxiv.1503.02531 · Cited by 14,204 · cited 1×
“[26, 58], quantization [118, 153], factorized embedding parameterization and cross-layer parameter sharing [70], and progressive module replacing [146].”
Deep Contextualized Word Representations
Peters, M., Neumann, M., Iyyer, M., Gardner, M., Clark, C., Lee, K., & Zettlemoyer, L. (2018) · Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers) · doi:10.18653/v1/n18-1202 · Cited by 11,983
Proceedings of the 24th international conference on Machine learning
(2007) · doi:10.1145/1273496 · Cited by 11,630
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
The Mathematical Theory of Communication
Shannon, C. E., Weaver, W., & Wiener, N. (1950) · Physics Today · doi:10.1063/1.3067010 · Cited by 9,801
Research works printed but not linked: 48
- [4] American Psychological Association. 2019. Discrimination: What it is, and how to cope. https://www.apa.org/topics/discrimination (2019). Google Scholar
- 10.18653/v1 · DOI
- 10.18653/v1 · DOI
- [15] Ruha Benjamin. 2019. Race After Technology: Abolitionist Tools for the New Jim Code. Polity Press, Cambridge, UK. Crossref Google Scholar
- [16] Elettra Bietti and Roxana Vatanparast. 2020. Data Waste. Harvard International Law Journal 61 (2020). Google Scholar
- 10.18653/v1 · DOI
- 10.18653/v1 · DOI
- 10.1016/0010-0277(86) · doi:10.1016/0010-0277(86
- [38] Terrance de Vries, Ishan Misra, Changhan Wang, and Laurens van der Maaten. 2019. Does object recognition work for everyone?. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops. 52--59. Google Scholar
- 10.18653/v1 · DOI
- [42] Ethan Fast, Tina Vachovsky, and Michael Bernstein. 2016. Shirtless and Dangerous: Quantifying Linguistic Signals of Gender Bias in an Online Fiction Writing Community. In Proceedings of the International AAAI Conference on Web and Social Media, Vol. 10. Google Scholar
- 10.18653/v1 · DOI
- [45] Darja Fišer, Ruihong Huang, Vinodkumar Prabhakaran, Rob Voigt, Zeerak Waseem, and Jacqueline Wernimont (Eds.). 2018. Proceedings of the 2nd Workshop on Abusive Language Online (ALW2). Association for Computational Linguistics, Brussels, Belgium. https://www.aclweb.org/anthology/W18-5100 Google Scholar
- [50] Batya Friedman, Peter H. Kahn, Jr., and Alan Borning. 2006. Value sensitive design and information systems. In Human-Computer Interaction in Management Information Systems: Foundations, P Zhang and D Galletta (Eds.). M. E. Sharpe, Armonk NY, 348--372. Google Scholar
- 10.18653/v1 · DOI
- 10.18653/v1 · DOI
- 10.18653/v1 · DOI
- [67] Brendan Kennedy, Drew Kogon, Kris Coombs, Joseph Hoover, Christina Park, Gwenyth Portillo-Wightman, Aida Mostafazadeh Davani, Mohammad Atari, and Morteza Dehghani. 2018. A typology and coding manual for the study of hate-based rhetoric. PsyArXiv. July 18 (2018). Google Scholar
- [77] Marianna Martindale and Marine Carpuat. 2018. Fluency Over Adequacy: A Pilot Study in Measuring User Trust in Imperfect MT. In Proceedings of the 13th Conference of the Association for Machine Translation in the Americas (Volume 1: Research Track). Association for Machine Translation in the Americas, Boston, MA, 13--25. https://www.aclweb.org/anthology/W18-1803 Google Scholar
- [81] Douglas M McLeod. 2007. News coverage and social protest: How the media's protect paradigm exacerbates social conflict. Journal of Dispute Resolution (2007), 185. Google Scholar
- [87] Robert C. Moore and William Lewis. 2010. Intelligent Selection of Language Model Training Data. In Proceedings of the ACL 2010 Conference Short Papers. Association for Computational Linguistics, Uppsala, Sweden, 220--224. https://www.aclweb.org/anthology/P10-2041 Google Scholar
- 10.18653/v1 · DOI
- [92] Maggie Nelson. 2015. The Argonauts. Graywolf Press, Minneapolis. Google Scholar
- 10.18653/v1 · DOI
- [96] David Ortiz, Daniel Myers, Eugene Walls, and Maria-Elena Diaz. 2005. Where do we stand with newspaper data? Mobilization: An International Quarterly 10, 3 (2005), 397--419. Google Scholar
- 10.3115/v1 · doi:10.3115/v1
- 10.18653/v1 · DOI
- 10.18653/v1 · DOI
- [106] Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever. 2019. Language models are unsupervised multitask learners. OpenAI Blog 1, 8 (2019), 9. Google Scholar
- 10.18653/v1 · DOI
- [109] Sarah T. Roberts, Joel Tetreault, Vinodkumar Prabhakaran, and Zeerak Waseem (Eds.). 2019. Proceedings of the Third Workshop on Abusive Language Online. Association for Computational Linguistics, Florence, Italy. https://www.aclweb.org/anthology/W19-3500 Google Scholar
- [112] Corby Rosset. 2020. Turing-NLG: A 17-billion-parameter language model by Microsoft. Microsoft Blog (2020). Google Scholar
- 10.18653/v1 · DOI
- 10.18653/v1 · DOI
- [121] Joonbo Shin, Yoonhyung Lee, and Kyomin Jung. 2019. Effective Sentence Scoring Method Using BERT for Speech Recognition. In Asian Conference on Machine Learning. 1081--1093. Google Scholar
- [124] Karen Spärck Jones. 2004. Language modelling's generative model: Is it rational? Technical Report. Computer Laboratory, University of Cambridge. Google Scholar
- [127] Katrina Srigley and Lorraine Sutherland. 2019. Decolonizing, Indigenizing, and Learning Biskaaybiiyang in the Field: Our Oral History Journey1. The Oral History Review (2019). Google Scholar
- 10.18653/v1 · DOI
- 10.18653/v1 · DOI
- 10.18653/v1 · DOI
- 10.18653/v1 · DOI
- 10.18653/v1 · DOI
- 10.18653/v1 · DOI
- 10.18653/v1 · DOI
- [150] Zhilin Yang, Zihang Dai, Yiming Yang, Jaime Carbonell, Russ R Salakhutdinov, and Quoc V Le. 2019. Xlnet: Generalized autoregressive pretraining for language understanding. In Advances in neural information processing systems. 5753--5763. Google Scholar
- 10.18653/v1 · DOI
- 10.18653/v1 · DOI
- 10.18653/v1 · DOI
Cited by
Items 1,601–1,700 of 6,931
Evaluating and Mitigating Discrimination in Language Model Decisions
Tamkin, A., Askell, A., Lovitt, L., Durmus, E., Joseph, N., Kravec, S., Nguyen, K., Kaplan, J., & Ganguli, D. (2023) · arXiv · doi:10.48550/arxiv.2312.03689 · Cited by 11
Can AI Be a Good Peer Reviewer? A Survey of Peer Review Process, Evaluation, and the Future
Wu, S., Jiang, O., Zhao, Y., Hu, T., Ma, Y., Zhang, K., Patwardhan, M., & Cohan, A. (2026) · Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/2026.acl-long.1504 · Cited by 11
Fully Autonomous AI Agents Should Not be Developed
Mitchell, M., Ghosh, A., Luccioni, A. S., & Pistilli, G. (2025) · arXiv · doi:10.48550/arxiv.2502.02649 · Cited by 11
Large Language Models (LLMs) and Generative Artificial Intelligence (GenAI)
Lee, R. (2025) · Natural Language Processing · doi:10.1007/978-981-96-3208-4_10 · Cited by 14
Environmental impacts of artificial intelligence in health care: considerations and recommendations
Richie, C., Hinrichs, S., Dobbe, R., French, P. J., Wei, J. C. J., Diehl, J. C., & Kong, R. (2025) · Health and Technology · doi:10.1007/s12553-025-01003-4 · Cited by 17
“And, as highlighted above, while articles on the carbon impact of health care [27] and the carbon impact of AI [28]are proliferating, the intersection of the three is overlooked.”
Liu, C., Zheng, J., Liu, Y., Wang, X., Zhang, Y., Yu, W., Yu, W., Yu, T., Jiang, W., Wang, D., et al. (2025) · International Journal for Equity in Health · doi:10.1186/s12939-025-02581-5 · Cited by 18
“However, because LLMs are typically trained on vast, uncurated, human-generated corpora, they tend to inter-nalize and replicate the prejudices, misrepresentation, derogatory language, and other discriminatory tendencies present in their training data [10, 11].”
Guenduez, A. A., Fuchs, S., Mergel, I. A., Mettler, T., & Frowein, S. C. J. (2025) · Quality & Quantity · doi:10.1007/s11135-025-02357-7 · Cited by 27
“Like any data-driven approach, automated text analysis can be susceptible to biases or stereotypes, as they are designed by humans and learn patterns from human-generated data (Bender et al. 2021; Macanovic2022).”
Development and validation of Generative AI Competence Scale (GenAIComp) among university students
Lee, S. C., Baby, T., Vongvit, R., Lee, J., Kim, Y., Cha, M. C., & Yoon, S. H. (2025) · Technology in Society · doi:10.1016/j.techsoc.2025.103059 · Cited by 33
Sui, C., Chang, C., & Yen, M.-H. (2025) · Journal of Science Education and Technology · doi:10.1007/s10956-025-10257-6 · Cited by 18
“GenAI responds with a high degree of confidence, but these are based on probabilistic models (Bender et al., 2021).”
Politics of the prompt: Government in the age of generative AI
Amoore, L., Bennett, S., Campolo, A., Jacobsen, B. N., & Rella, L. (2025) · Economy and Society · doi:10.1080/03085147.2025.2560177 · Cited by 16
Stofiana, T., Sunendar, D., Mulyati, Y., & Sastromiharjo, A. (2025) · Ampersand · doi:10.1016/j.amper.2025.100242 · Cited by 28
Vinay, R., Уэтова, Е., Tommila, N. C., Biller‐Andorno, N., & Kowatsch, T. (2025) · JMIR Aging · doi:10.2196/76489 · Cited by 14
“LLMs’ reliance on large-scale computational resources raises challenges related to latency, particularly in real-time applications [56], hallucinations, that is, LLMs can generate nonsensical, incorrect text [57], unpredictable and unexpected responses [52], and bias [58,59].”
Jones, K. M. L. (2025) · International Journal of Qualitative Methods · doi:10.1177/16094069251404329 · Cited by 32
“Researchers may gain access to automated coding and content analysis features without visibility into the model ’ s architecture or training corpus, and the biases embedded therein (Bender et al., 2021).”
Language machines: Toward a linguistic anthropology of large language models
Lamoureaux, S., Castelle, M., & Weichselbraun, A. (2025) · Journal of Linguistic Anthropology · doi:10.1111/jola.70033 · Cited by 20
Kim, Y., Shin, S., & Yoo, S.-H. (2025) · BMC Medical Ethics · doi:10.1186/s12910-025-01316-z · Cited by 18
Hamamra, B., Khlaif, Z. N., & Mahamid, N. (2026) · International Journal of Educational Technology in Higher Education · doi:10.1186/s41239-026-00597-7 · Cited by 26
“Bender et al. (2021) warn that large language models can produce fluent language without grounded understanding, while Kasneci et al. (2023) identify both the educational opportunities and the risks of generative AI in learning contexts.”
Domesticating Artificial Intelligence
Müller, L. K. (2022) · Moral Philosophy and Politics · doi:10.1515/mopp-2020-0054 · Cited by 16
Majdik, Z. P., & Wynn, J. (2022) · Technical Communication Quarterly · doi:10.1080/10572252.2022.2077452 · Cited by 14
“As we discussed earlier, scholarship by Bender et al. (2021), Strubell et al. (2019), and others have identified CO 2 emissions, equity/access issues stemming from the computing cost needed to use models like BERT, and bias built into large training data sets, as challenges necessary to address.”
BackgroundMethodHeyn, H.-M., Habibullah, K. M., Knauss, E., Horkoff, J., Borg, M., Knauss, A., & Li, P. J. (2023) · IEEE/ACM International Conference on AI Engineering – Software Engineering for AI (CAIN) · doi:10.1109/cain58948.2023.00011 · Cited by 23
Assessing the Prevalence and Predictors of Incivility in Online News Comments Across Six Countries
Salgado, S., de Zúñiga, H. G., da Silva, P. A., Biscaia, A., Coimbra, M. E., Martins, B., & Francisco, A. P. (2023) · Journalism Practice · doi:10.1080/17512786.2023.2246431 · Cited by 16
“We note that previous literature has pointed to issues in the use of large language models for text classi fi cation, for instance Bender et al. (2021) suggested that these models, in addition to involving signi fi cant computational resources and energy consumption, can encode hegemonic views and…”
MethodRevolutionizing Formative Assessment in STEM Fields: Leveraging AI and NLP Techniques
Tan, C. W., & Lim, K. Y. (2023) · Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) · doi:10.1109/apsipaasc58517.2023.10317226 · Cited by 19
“LLMs represent a major advancement in AI, with the promise of transforming domains through learned knowledge [21].”
BackgroundDefinitions Matter: Guiding GPT for Multi-label Classification
Peskine, Y., Korenčić, D., Grubišić, I., Papotti, P., Troncy, R., & Rosso, P. (2023) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2023.findings-emnlp.267 · Cited by 47
“Biases We are aware of the biases of LLMs in classification tasks (Bender et al., 2021).”
Bouyamourn, A. Z. A. (2023) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2023.emnlp-main.192 · Cited by 33
Culturally Aware Natural Language Inference
Huang, J., & Yang, D. (2023) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2023.findings-emnlp.509 · Cited by 98
“LLMs generated contents may contain artifacts, societal biases, and toxic texts (Lucy and Bamman, 2021; Bender et al., 2021; Sheng et al., 2021; Yuan et al., 2021; Borchers et al., 2022).”
BackgroundIs ChatGPT the ultimate Data Augmentation Algorithm?
Piedboeuf, F., & Langlais, P. (2023) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2023.findings-emnlp.1044 · Cited by 39
“…come with a plethora of ethical problems which have been well discussed in the litterature, including the environmental cost (Schwartz et al., 2019; Bender et al., 2021) and environmental racism (Rillig et al., 2023), the repetition of learned biases against minorities (Singh, 2023), and…”
BackgroundKnowledge and Digital Technology
Glückler, J., & Panitz, R. (2024) · Knowledge and space · doi:10.1007/978-3-031-39101-9 · Cited by 13
Behizadeh, N., Johnson, L. L., & Garcia, M. (2023) · English Education · doi:10.58680/ee20235618 · Cited by 10
From sci-fi to the classroom: Implications of AI in task-based writing
Poole, F. J., & Polio, C. G. (2023) · TASK Journal on Task-Based Language Teaching and Learning · doi:10.1075/task.00022.poo · Cited by 14
Biden’s Executive Order on AI: strengths, weaknesses, and possible reform steps
Wörsdörfer, M. (2024) · AI and Ethics · doi:10.1007/s43681-024-00510-w · Cited by 14
“Critics argue that these technologies have the potential to exacerbate issues such as disinformation, deception, fraud, manipulation, and mistrust [26–30].”
Wandera, D. B. (2024) · Reading Research Quarterly · doi:10.1002/rrq.572 · Cited by 22
An Agile Pathway Towards Carbon-Aware Clouds
Patel, P., Gregersen, T., & Anderson, T. E. (2024) · ACM SIGEnergy Energy Informatics Review · doi:10.1145/3698365.3698368 · Cited by 12
Introducing ChatSQC: Enhancing statistical quality control with augmented AI
Megahed, F. M., Chen, Y., Zwetsloot, I. M., Knoth, S., Montgomery, D. C., & Jones‐Farmer, L. A. (2024) · Journal of Quality Technology · doi:10.1080/00224065.2024.2372328 · Cited by 29
“For example, with the development of Othello-GPT (Li et al. 2023), the authors showed that LLMs are not just “stochastic parrots” (Bender et al. 2021); instead, LLMs can create a correct model related to the Othello game by being trained on the moves and not the rules of the game.”
MethodThe Elusiveness of Detecting Political Bias in Language Models
Lunardi, R., Barbera, D. L., & Roitero, K. (2024) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3627673.3680002 · Cited by 30
“Among various biases like those based on race, gender, or religion [1, 3, 9, 15], political bias in LLMs is particularly concerning, as it can subtly influence public opinion, reinforce societal stereotypes, and raise societal decision-making issues in sensitive domains.”
The Cultivated Practices of Text-to-Image Generation
Oppenlaender, J. (2024) · Springer eBooks · doi:10.1007/978-3-031-66528-8_14 · Cited by 18
Kabashkin, I. (2024) · Mathematics · doi:10.3390/math12213428 · Cited by 13
A comprehensive analysis of gender, racial, and prompt-induced biases in large language models
Torres, N., Ulloa, C., Araya, I., Ayala, M., & Jara, S. (2024) · International Journal of Data Science and Analytics · doi:10.1007/s41060-024-00696-6 · Cited by 20
The Three Social Dimensions of Chatbot Technology
Figueroa-Torres, M. (2024) · Philosophy & Technology · doi:10.1007/s13347-024-00826-9 · Cited by 17
“This progression has also sparked critical discussions on the technical limitations and flaws, exemplified by Emily Bender’s et al. coining of the term stochastic parrots to describe the superficiality of these models by merely predicting sequential word probabilities (Bender et al., 2021).”
Exploring ChatGPT’s role in English grammar learning: A Kolb model perspective
Annamalai, N., & Bervell, B. (2025) · Innovations in Education and Teaching International · doi:10.1080/14703297.2025.2451328 · Cited by 20
The private sector is hoarding AI researchers: what implications for science?
Jurowetzki, R., Hain, D. S., Wirtz, K., & Bianchini, S. (2025) · AI & Society · doi:10.1007/s00146-024-02171-z · Cited by 21
“Gebru’s paper, in which she and coau-thors raised some concerns about the inherent biases and resource consumption of large language models (Bender et al. 2021), led to a dispute with management and ultimately her termination.”
Bossert, L. N., & Loh, W. (2025) · Nature Machine Intelligence · doi:10.1038/s42256-025-00979-y · Cited by 19
Risks, Causes, and Mitigations of Widespread Deployments of Large Language Models (LLMs): A Survey
Sakib, M. N., Islam, M. A., Pathak, R., & Arifin, M. M. (2024) · International Conference on Artificial Intelligence, Blockchain, and Internet of Things (AIBThings) · doi:10.1109/aibthings63359.2024.10863356 · Cited by 30
Generative AI Under Scrutiny: Assessing the Risks and Challenges in Diverse Domains
Nidhisree, C., Paul, A., Venunadh, A., & Bhowmick, R. S. (2024) · IEEE International Conference on Cybernetics, Cognition and Machine Learning Applications (ICCCMLA) · doi:10.1109/icccmla63077.2024.10871350 · Cited by 19
“Prioritizing these impacts within the research community [3] can be achieved by evaluating costs based on resource consumption. e) Marginalization impact: Large datasets sourced from the internet often over-represent dominant viewpoints and biases that can harm marginalized populations.”
InfluentialTowards Visual Analytics for Explainable AI in Industrial Applications
Kucher, K., Zohrevandi, E., & Westin, C. (2025) · Analytics · doi:10.3390/analytics4010007 · Cited by 17
Moore, B., Herington, J., & Tekin, Ş. (2025) · The Journal of Pediatrics · doi:10.1016/j.jpeds.2025.114509 · Cited by 11
Ethics of AI: Toward a “Design for Values” Approach
Buijsman, S., Klenk, M., & van den Hoven, J. (2025) · Cambridge University Press eBooks · doi:10.1017/9781009367783.005 · Cited by 20
Metaethical perspectives on ‘benchmarking’ AI ethics
LaCroix, T., & Luccioni, A. (2022) · AI and Ethics · doi:10.1007/s43681-025-00703-x · Cited by 23
Materiality and risk in the age of pervasive AI sensors
Sloane, M., Moss, E., Kennedy, S., Stewart, M. P., Warden, P., Plancher, B., & Reddi, V. J. (2025) · Nature Machine Intelligence · doi:10.1038/s42256-025-01017-7 · Cited by 19
Cellard, L., Parker, C., & Haines, F. (2025) · Environment and Planning E Nature and Space · doi:10.1177/25148486251332087 · Cited by 44
“…raising a series of wide-ranging ethical questions about Google ’ s language processing tools, focused mostly on encoded bias, but also highlighting the energy use of machine learning and asking whether that resource could be more bene fi cially used elsewhere (Bender et al., 2021; Tiku, 2020).”
LLM Whisperer: An Inconspicuous Attack to Bias LLM Responses
Lin, W., Gerchanovsky, A., Akgul, O., Bauer, L., Fredrikson, M., & Wang, Z. (2025) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3714025 · Cited by 28
“The societal topic bias we explore in this paper (e.g., countries, political parties, candidates) remains relatively under-explored.”
A Critical Analysis of Machine Learning Eco-feedback Tools through the Lens of Sustainable HCI
Görücü, S., Morais, L., & Panagiotidou, G. (2025) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3713198 · Cited by 17
From intelligence to autopoiesis: rethinking artificial intelligence through systems theory
Zönnchen, B., Dzhimova, M., & Socher, G. (2025) · Frontiers in Communication · doi:10.3389/fcomm.2025.1585321 · Cited by 21
Understanding Social Biases in Large Language Models
Gupta, O., Marrone, S., Gargiulo, F., Jaiswal, R. R., & Marassi, L. (2025) · AI · doi:10.3390/ai6050106 · Cited by 22
Sphinteract: Resolving Ambiguities in NL2SQL through User Interaction
Zhao, F., Deep, S., Psallidas, F., Floratou, A., Agrawal, D., & Abbadi, A. E. (2024) · Proceedings of the VLDB Endowment · doi:10.14778/3717755.3717772 · Cited by 23
“Second, the CQs algorithm has the disadvantage that LLMs can continue to generate clarification questions (owing to their stochastic parrot [5] nature where LLMs may continue to generate output without truly understanding the task) even if all ambiguities have been resolved.”
Rodrigues, T. V. (2025) · The International Journal of Social Sciences and Humanities Invention · doi:10.18535/ijsshi/v12i05.04 · Cited by 11
“As Bender et al. (2021) emphasize, LLMs do not understand the texts they produce; they do not have beliefs, goals, or semantic comprehension.”
State Space Models as Foundation Models: A Control Theoretic Overview
Alonso, C. A., Sieber, J., & Zeilinger, M. N. (2025) · Proceedings of the ... American Control Conference/Proceedings of the American Control Conference · doi:10.23919/acc63710.2025.11107969 · Cited by 53
Jung, T., & Joe, I. (2025) · Applied Sciences · doi:10.3390/app15179398 · Cited by 10
Tierens, T., Decuypere, M., Hartong, S., & Alirezabeigi, S. (2025) · Postdigital Science and Education · doi:10.1007/s42438-025-00585-6 · Cited by 17
“…high-lighting how technologies like ChatGPT reanimate political in-/visibilities and in-/equalities (Gillespie 2024), parrot normative assumptions to users (Bender et al. 2021), and lay foundations for different forms of educational practice and sub-jectivity to emerge (Perrotta et al. 2024).”
Unequal Opportunities: Examining the Bias in Geographical Recommendations by Large Language Models
Dudy, S., Tholeti, T., Ramachandranpillai, R., Ali, M., Li, T. J.-J., & Baeza-Yates, R. A. (2025) · ACM International Conference on Intelligent User Interfaces (IUI) · doi:10.1145/3708359.3712111 · Cited by 30
Liu, F., Farkiani, B., & Crowley, P. (2025) · Computer Networks · doi:10.1016/j.comnet.2025.111614 · Cited by 29
Heidenreich, H. S., & Williams, J. R. (2021) · Proceedings of the AAAI/ACM Conference on AI Ethics and Society · doi:10.1145/3461702.3462578 · Cited by 13
“In light of all these aspects observed, we encourage readers to further reflect on the trend of ever-increasing model and dataset size [2], especially without sufficient documentation (e.g. Datasheets [10] or Model Cards [18]) and careful consideration of how a model will be used and by whom.”
Intersectional Inquiry, on the Ground and in the Algorithm
Robertson, S., Magee, L., & Soldatić, K. (2022) · Qualitative Inquiry · doi:10.1177/10778004221099560 · Cited by 15
Combining Transformer Embeddings with Linguistic Features for Complex Word Identification
Zambrano, J. O., Espín-Riofrio, C., & Montejo‐Ráez, A. (2022) · Electronics · doi:10.3390/electronics12010120 · Cited by 14
FairDistillation: Mitigating Stereotyping in Language Models
Delobelle, P., & Berendt, B. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-26390-3_37 · Cited by 32
AI-Aided Ceramic Sculptures: Bridging Deep Learning with Materiality
Guljajeva, V., & Solà, M. C. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-29956-8_23 · Cited by 11
“This phenomenon is related to Bender and Gebru’s findings of language models [4] and is associated with the complexity of dealing with dimensional spaces of mean-368 ing.”
AŞKUN, V. (2023) · Erciyes Akademi · doi:10.48070/erciyesakademi.1281544 · Cited by 15
Generation-Based Data Augmentation for Offensive Language Detection: Is It Worth It?
Casula, C., & Tonelli, S. (2023) · Conference of the European Chapter of the Association for Computational Linguistics · doi:10.18653/v1/2023.eacl-main.244 · Cited by 11
“Even though large language models have been found to have a tendency to encode and propagate undesirable social bias (Bender et al., 2021), going as far as generating toxic sequences starting from non-toxic prompts (Gehman et al., 2020), the use of synthetic data for offensive language detection…”
BackgroundScreens, streams, and flows: Implications of digital platforms for aquatic citizen science
Millar, E., Melles, S., & Rinner, C. (2023) · Geoforum · doi:10.1016/j.geoforum.2023.103864 · Cited by 11
On Political Theory and Large Language Models
Rodman, E. (2023) · Political Theory · doi:10.1177/00905917231200826 · Cited by 17
“…related to unjust distributions of harms from these models, and that is the widely acknowledged (Guo and Caliskan 2020) presence of biases and stereotypes: the models “encode stereotypical and derogatory associations along gender, race, ethnicity, and disability status” (Bender et al. 2021, 613).”
BackgroundTekumalla, R., & Banda, J. M. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-48044-7_26 · Cited by 22
From Vulnerabilities to Improvements- A Deep Dive into Adversarial Testing of AI Models
Hannon, B., Kumar, Y., Sorial, P., Li, J. J., & Morreale, P. A. (2023) · Congress in Computer Science, Computer Engineering, and Applied Computing (CSCE) · doi:10.1109/csce60160.2023.00422 · Cited by 15
Tang, Y., & Ha, D. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2109.02869 · Cited by 84
The pitfalls of (non-definitive) Environmental, Social, and Governance scoring methodologyPreprint
Sahin, Ö., Bax, K., Paterlini, S., et al. (2022) · SSRN Electronic Journal · doi:10.2139/ssrn.4020354 · Cited by 10
Beyond the hype: large language models propagate race-based medicinePreprint
Omiye, J. A., Lester, J., Spichak, S., Rotemberg, V. M., & Daneshjou, R. (2023) · medRxiv · doi:10.1101/2023.07.03.23292192 · Cited by 10
“LLMs have been known to also generate nonsensical responses 24,25 ; while this study did not systematically assess these, we noted that equations generated by the models were fabricated.”
Fast-Fnet: Accelerating Transformer Encoder Models Via Efficient Fourier LayersPreprint
Sevim, N., Özyedek, E. O., Şahinuç, F., & Koç, A. (2023) · SSRN Electronic Journal · doi:10.2139/ssrn.4566618 · Cited by 12
Generating Inflation Expectations with Large Language ModelsPreprint
Zarifhonarvar, A. (2024) · SSRN Electronic Journal · doi:10.2139/ssrn.4825076 · Cited by 10
The limitations of large language models for understanding human language and cognitionPreprint
Cuskley, C. F., Woods, R. J. H., & Flaherty, M. (2024) · doi:10.31234/osf.io/hbwrg · Cited by 10
Computational Hermeneutics: Evaluating Generative AI as a Cultural TechnologyPreprint
Kommers, C., Ahnert, R., Antoniak, M., Benetos, E., Benford, S., Bunz, M., Caramiaux, B., Concannon, S. J., Disley, M., Dobson, J. E., et al. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5409144 · Cited by 10
Neither Valid nor Reliable? Investigating the Use of LLMs as Judges
Chehbouni, K., Haddou, M., Cheung, J. C., & Farnadi, G. (2025) · Advances in Neural Information Processing Systems 38 · doi:10.52202/085713-3029 · Cited by 10
Thieme, A., Rajamohan, A., Cooper, B., Groombridge, H., Simister, R., Wong, B., Woznitza, N., Pinnock, M. A., Wetscherek, M. T., Morrison, C., et al. (2024) · arXiv · doi:10.48550/arxiv.2405.05299 · Cited by 10
Judgments of learning distinguish humans from large language models in predicting memory
Huff, M., & Ulakci, E. (2025) · Scientific Reports · doi:10.1038/s41598-025-22290-x · Cited by 9
Fathy, E. A., Zidan, H. A. K. Y., Abdou, A. H. E. S. A., & Fouad, A. M. (2025) · Tourism and Hospitality Research · doi:10.1177/14673584251357510 · Cited by 17
AI in Automated and Remote UX Evaluation: A Systematic Review (2014–2024)
Liu, J. (2025) · Advances in Human-Computer Interaction · doi:10.1155/ahci/7442179 · Cited by 22
“Risks, including hallucinated results, overgeneralization, and dif-fculty interpreting outputs, may mislead UX decision-making if not rigorously validated [9, 10].”
The quasi-domestication of social chatbots: The case of Replika
Neff, G., & Nagy, P. (2025) · New Media & Society · doi:10.1177/14614448251359218 · Cited by 16
“Rather than tackling these important issues, however, tech companies – just like Luka–largely ignore them (e.g. Bender et al., 2021; Dwork and Minow, 2022).”
A Network Approach to Public Trust in Generative AI
McIntyre, A., Conover, L., & Russo, F. (2025) · Philosophy & Technology · doi:10.1007/s13347-025-00974-6 · Cited by 9
“Certainly, these technologies lack any conscious awareness of the meaning of their outputs in the same way as humans and have been rightly described as “stochastic parrots” (Bender et al., 2021).”
Leveraging Large Language Models for Real-Time UAV Control
Choutri, K., Fadloun, S., Khettabi, A., Lagha, M., Meshoul, S., & Fareh, R. (2025) · Electronics · doi:10.3390/electronics14214312 · Cited by 13
Viljakainen, A., Silmukari, J., Roto, V., Čaić, M., & Aromaa, S. (2025) · Behaviour and Information Technology · doi:10.1080/0144929x.2025.2596889 · Cited by 17
Beyond overconfidence: Embedding curiosity and humility for ethical medical AI
Cajas, S. A., Castro, R., Celi, L. A., Reyes, R. D., Engelmann, J., Ercole, A., Hilel, A., Kalla, M., Kinyera, L., Lange, M., et al. (2026) · PLOS Digital Health · doi:10.1371/journal.pdig.0001013 · Cited by 17
Liu, H., Fang, X., Cui, Q., & Chiu, T. K. F. (2026) · Review of Education · doi:10.1002/rev3.70141 · Cited by 14
Hodzic, S., Stević, A., & Matthes, J. (2026) · Technology in Society · doi:10.1016/j.techsoc.2026.103331 · Cited by 12
Into the Unknown: Exploration of Search Engines’ Responses to Users with Depression and Anxiety
Milton, A., & Pera, M. S. (2023) · ACM Transactions on the Web · doi:10.1145/3580283 · Cited by 10
Bias in Pruned Vision Models: In-Depth Analysis and Countermeasures
Iofinova, E., Peşte, A., & Alistarh, D. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.02334 · Cited by 21
Regulation and NLP (RegNLP): Taming Large Language Models
Goanță, C., Αλέτρας, Ν., Chalkidis, I., Ranchordás, S., & Spanakis, G. (2023) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2023.emnlp-main.539 · Cited by 29
Bevara, R. V. K., Mannuru, N. R., Karedla, S. P., & Xiao, T. (2024) · Applied Sciences · doi:10.3390/app14083483 · Cited by 15
“Base models serve as a reference point for grasping the inherent bias present in transformer models at the outset of the investigation (Bender et al. [32]; Caliskan et al. [16]).”
BackgroundLarge language models and the future of academic writing
Nayak, P. R., & Gogtay, N. (2024) · Journal of Postgraduate Medicine · doi:10.4103/jpgm.jpgm_823_23 · Cited by 9
“[3] Apart from democratizing access to information and analyses of factual data, we literally have the world of English language at our fingertips.”
AI-Driven Support for People with Speech & Language Difficulties
Dangol, A., Huang, Y., Setlur, S., Smolansky, A., Subramonyam, H., Suh, H., Xiong, J., & Kientz, J. A. (2024) · CHI Conference Extended Abstracts on Human Factors in Computing Systems (CHI EA) · doi:10.1145/3613905.3643984 · Cited by 14
They Prefer Humans! Experimental Measurement of Student Trust in ChatGPT
Walker, F., Favetta, M., Hasker, L., & Walker, R. D. (2024) · CHI Conference Extended Abstracts on Human Factors in Computing Systems (CHI EA) · doi:10.1145/3613905.3650955 · Cited by 21
Algorithmic Issues, Challenges, and Theoretical Concerns of ChatGPT
Patil, P., Kulkarni, K., & Sharma, P. (2024) · Advances in computational intelligence and robotics book series · doi:10.4018/979-8-3693-6824-4.ch003 · Cited by 13
Hoi, S., Yu, Y., & Ye, L. (2023) · International Conference on Information Education and Artificial Intelligence (ICIEAI) · doi:10.1145/3660043.3660182 · Cited by 10
Post-Rhetoric: A Rhetorical Profile of the Generative Artificial Intelligence Chatbot
Wang, Z. (2024) · Rhetoric Review · doi:10.1080/07350198.2024.2351723 · Cited by 18
Arpin, M.-L., Leclerc, S. H., & Lonca, G. (2024) · Sustainability · doi:10.3390/su16156364 · Cited by 11
Metadata
- Authors
- Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S.
- Affiliations
- University of Washington (US)
- Black in AI, Palo Alto, CA, USA
- The Aether
- Venue
- Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency
- Publisher
- ACM
- Published
- 2021-03
- Pages
- 610-623
- OpenAlex
- W3133702157
- Cited by
- 6,988
- References
- 117
- Citations by year
- 2026: 2484 · 2025: 1887 · 2024: 1153 · 2023: 842 · 2022: 384 · 2021: 229
- FWCI
- 300.72
- Type
- conference paper
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- Locations
- ACM Conference on Fairness, Accountability, and Transparency (FAccT) (publishedVersion, cc-by)
- Creighton Digital Repository (CDR) (submittedVersion)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
- Physical Sciences › Computer Science › Artificial Intelligence › Natural Language Processing Techniques
- Keywords
- large language models, model size, AI risks, financial costs, data curation, dataset documentation, stakeholder value, BERT, GPT, fine-tuning
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.
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.