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 1,001–1,100 of 6,931
Heads we win, tails you lose: AI detectors in education
Bassett, M. A., Bradshaw, W., Bornsztejn, H., Hogg, A., Murdoch, K., Pearce, B., & Webber, C. (2026) · Journal of Higher Education Policy and Management · doi:10.1080/1360080x.2026.2622146 · Cited by 24
van Duijn, M., van Dijk, B., Kouwenhoven, T., de Valk, W., Spruit, M., & vanderPutten, P. (2023) · Proceedings of the 27th Conference on Computational Natural Language Learning (CoNLL) · doi:10.18653/v1/2023.conll-1.25 · Cited by 24
Viduani, A. C., Cosenza, V., Fisher, H. L., Buchweitz, C., Piccin, J., Pereira, R. B., Kohrt, B. A., Mondelli, V., van Heerden, A. C., Araújo, R., et al. (2023) · JMIR Human Factors · doi:10.2196/44388 · Cited by 30
“In addition, we believe that, as a tool, the IDEABot supports stakeholder values [47].”
MethodArtificial Design Fiction: Using AI as a Material for Pastiche Scenarios
Blythe, M. (2023) · International Academic Mindtrek Conference (Mindtrek) · doi:10.1145/3616961.3616987 · Cited by 35
The impacts of AI futurism: an unfiltered look at AI's true effects on the climate crisis
Schütze, P. (2024) · Ethics and Information Technology · doi:10.1007/s10676-024-09758-6 · Cited by 38
““Is it fair … that the residents of the Maldives (likely to be underwater by 2100) or the 800,000 people in Sudan affected by drastic floods pay the environmental price of training and deploying ever larger English LMs”, such as ChatGPT (Bender et al., 2021, p. 612)?”
Think Fast, Think Slow, Think Critical: Designing an Automated Propaganda Detection Tool
Zavolokina, L., Sprenkamp, K., Katashinskaya, Z., Jones, D. G., & Schwabe, G. (2024) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642805 · Cited by 52
On Cyber-Social Learning: A Critique of Artificial Intelligence in Education
Cope, B., & Kalantzis, M. (2024) · Postdigital science and education · doi:10.1007/978-3-031-64487-0_1 · Cited by 50
Lee, J., Cha, H., Hwangbo, Y., & Cheon, W. (2024) · Journal of Personalized Medicine · doi:10.3390/jpm14121131 · Cited by 31
Torkamaan, H., Steinert, S., Pera, M. S., Kudina, O., Freire, S. K., Verma, H., Kelly, S., Sekwenz, M.-T., Yang, J., van Nunen, K., et al. (2024) · Behaviour and Information Technology · doi:10.1080/0144929x.2024.2431068 · Cited by 43
Secret Use of Large Language Model (LLM)
Zhang, Z., Shen, C., Yao, B., Wang, D., & Li, T. (2025) · Proceedings of the ACM on Human-Computer Interaction · doi:10.1145/3711061 · Cited by 48
de Almeida, J. G., Messiou, C., Withey, S. J., Matos, C., Koh, D.-M., & Papanikolaou, N. (2025) · European Radiology · doi:10.1007/s00330-025-11654-6 · Cited by 23
Research evaluation with ChatGPT: is it age, country, length, or field biased?
Thelwall, M. A., & Kurt, Z. (2025) · Scientometrics · doi:10.1007/s11192-025-05393-0 · Cited by 42
“Bias can occur even when medical guidelines are clear and have been ingested by a LLM because its non-expert training data also influences its output (e.g., Bender et al., 2021; Gururangan et al., 2020).”
Generative AI for computational chemistry: A roadmap to predicting emergent phenomena
Tiwary, P., Herron, L., John, R., Lee, S., Sanwal, D., & Wang, R.-Y. (2025) · Proceedings of the National Academy of Sciences of the United States of America · doi:10.1073/pnas.2415655121 · Cited by 26
Artificial Intelligence Project Success Factors—Beyond the Ethical Principles
Miller, G. J. (2022) · Lecture notes in business information processing · doi:10.1007/978-3-030-98997-2_4 · Cited by 41
Handling and Presenting Harmful Text in NLP Research
Kirk, H., Birhane, A., Vidgen, B., & Derczynski, L. (2022) · Findings of the Association for Computational Linguistics: EMNLP 2022 · doi:10.18653/v1/2022.findings-emnlp.35 · Cited by 25
“Harms in datasets can transfer to the models trained on them (Bender et al., 2021; Rauh et al., 2022).”
BackgroundMethods for Measuring, Updating, and Visualizing Factual Beliefs in Language Models
Hase, P., Diab, M. T., Celikyilmaz, A., Li, X., Kozareva, Z., Stoyanov, V., Bansal, M., & Iyer, S. (2023) · Conference of the European Chapter of the Association for Computational Linguistics · doi:10.18653/v1/2023.eacl-main.199 · Cited by 73
“…LMs (1) struggle with future data when trained on data from the past (Lazaridou et al., 2021; Dhingra et al., 2021), (2) often generate morally undesirable text (Gehman et al., 2020; Bender et al., 2021), and (3) simply give inaccurate outputs for tasks like question answering (Lin et al., 2021).”
BackgroundWill Code Remain a Relevant User Interface for End-User Programming with Generative AI Models?
Sarkar, A. (2023) · Proceedings of the 2023 ACM SIGPLAN International Symposium on New Ideas, New Paradigms, and Reflections on Programming and Software · doi:10.1145/3622758.3622882 · Cited by 23
Prediction of outcomes after cardiac arrest by a generative artificial intelligence model
Amacher, S. A., Arpagaus, A., Sahmer, C., Becker, C., Groß, S., Urben, T., Tisljar, K., Sutter, R., Marsch, S. C. U., & Hunziker, S. (2024) · Resuscitation Plus · doi:10.1016/j.resplu.2024.100587 · Cited by 27
Generation Probabilities Are Not Enough: Uncertainty Highlighting in AI Code Completions
Vasconcelos, H., Bansal, G., Fourney, A., Liao, Q. V., & Vaughan, J. W. (2024) · ACM Transactions on Computer-Human Interaction · doi:10.1145/3702320 · Cited by 91
Fringe News Networks: Dynamics of US News Viewership following the 2020 Presidential Election
KhudaBukhsh, A. R., Sarkar, R., Kamlet, M. S., & Mitchell, T. M. (2022) · ACM Web Science Conference (WebSci) · doi:10.1145/3501247.3531577 · Cited by 31
Generative AI and the Workforce: What Are the Risks?Preprint
Walkowiak, E., & MacDonald, T. (2023) · SSRN Electronic Journal · doi:10.2139/ssrn.4568684 · Cited by 32
Dwivedi, Y. K., Hughes, L., Bhadeshia, H. K. D. H., Ananiadou, S., Cohn, A. G., Cole, J. M., Conduit, G. J., Desarkar, M. S., & Wang, X. (2023) · International Journal of Information Management · doi:10.1016/j.ijinfomgt.2023.102725 · Cited by 29
Reframing deception for human-centered AI
Umbrello, S., & Natale, S. (2024) · International Journal of Social Robotics · doi:10.1007/s12369-024-01184-4 · Cited by 23
“…for processes like Z-Inspection®, not only permits but supports the creation of reliable, safe, and trustworthy CAI systems across all levels of abstraction 5 by targeting each level rather than a unitary top-down approach that would be, at best, ineffective if not outright impossible to achieve.”
Sun, Y., & Xu, C. (2026) · Journal of Knowledge Management · doi:10.1108/jkm-02-2025-0252 · Cited by 23
Generation Probabilities Are Not Enough: Uncertainty Highlighting in AI Code Completions
Vasconcelos, H., Bansal, G., Fourney, A., Liao, Q. V., & Vaughan, J. W. (2023) · arXiv · doi:10.48550/arxiv.2302.07248 · Cited by 23
An Investigation of Applying Large Language Models to Spoken Language Learning
Gao, Y., Nuchged, B., Li, Y., & Peng, L. (2023) · Applied Sciences · doi:10.3390/app14010224 · Cited by 42
Torkamaan, H., Tahaei, M., Buijsman, S., Xiao, Z., Wilkinson, D., & Knijnenburg, B. P. (2024) · Human-computer interaction series · doi:10.1007/978-3-031-55109-3_2 · Cited by 42
Maximising Large Language Model Utility in Cardiovascular Care: A Practical Guide
Nolin-Lapalme, A., Thériault-Lauzier, P., Corbin, D., Tastet, O., Sharma, A., Hussin, J., Kadoury, S., Jiang, R., Krahn, A. D., Gallo, R., et al. (2024) · Canadian Journal of Cardiology · doi:10.1016/j.cjca.2024.05.024 · Cited by 39
Rhetoric of/with AI: An Introduction
Majdik, Z. P., & Graham, S. S. (2024) · Rhetoric Society Quarterly · doi:10.1080/02773945.2024.2343264 · Cited by 32
Future of Education with Neuro-Symbolic AI Agents in Self-Improving Adaptive Instructional Systems
Tong, R. J., & Hu, X. (2024) · Frontiers of digital education. · doi:10.1007/s44366-024-0008-9 · Cited by 28
“They aim for coherence and relevance but can produce incorrect or nonsensical outputs (Bender et al., 2021).”
van den Broek, S., Sankaran, S., de Wit, J., & de Rooij, A. (2024) · Participatory Design Conference · doi:10.1145/3661455.3669868 · Cited by 39
“There are also general concerns regarding the biases inherent to (generative) AI systems [2, 13, 15, 43] and ownership of generated content by such systems [21].”
Walsh, J., Cave, J. A. K., & Griffiths, F. E. (2024) · Journal of Medical Internet Research · doi:10.2196/54321 · Cited by 27
“AI is Soulless”: Hollywood Film Workers’ Strike and Emerging Perceptions of Generative Cinema
Halperin, B. A., & Rosner, D. K. (2025) · ACM Transactions on Computer-Human Interaction · doi:10.1145/3716135 · Cited by 43
“More broadly, it may also hinge on reflexive [14] and accountable data science practices (e.g., smaller and documented datasets) [6] built for specific purposes with concern for many worlds of ethics [107] rather than universalized “Artificial General Intelligence” for mass automation.”
Synthetic media and computational capitalism: towards a critical theory of artificial intelligence
Berry, D. M. (2025) · AI & Society · doi:10.1007/s00146-025-02265-2 · Cited by 42
“This article contributes to key debates in contemporary scholarship (Nadin 2019; Bender et al. 2021; Coeckelbergh 2022; Mishra 2023; Vallor 2024; Gill 2024; see Coeckel-bergh and Gunkel 2024).”
Zhang, L., Yao, Z., & Moghaddam, A. H. (2025) · Journal of Teacher Education · doi:10.1177/00224871251325109 · Cited by 31
“Moreover, they primarily simulate human-like reasoning based on statistical patterns learned from their training data (Bender et al., 2021).”
Grab, M. Ö. (2025) · Innovative Higher Education · doi:10.1007/s10755-025-09801-4 · Cited by 29
Liang, H., Fan, J., & Wang, Y. (2025) · Sustainability · doi:10.3390/su17093842 · Cited by 44
Attard-Frost, B., & Widder, D. G. (2025) · Big Data & Society · doi:10.1177/20539517251340603 · Cited by 62
“Inaccurate predictions, decisions, and other data outputs created through the use of unreliable ML models cause many social, political, economic, physical, and psychological harms (Angwin et al., 2016; Bender et al., 2021; Grote and Berens, 2022; Mökander and Axente, 2023; Rankin et al., 2020).”
Prompt-to-SQL Injections in LLM-Integrated Web Applications: Risks and Defenses
Pedro, R., Coimbra, M. E., de Castro, D. A., Carreira, P., & Santos, N. (2025) · ACM/IEEE International Conference on Software Engineering: Companion Proceedings (ICSE-Companion) · doi:10.1109/icse55347.2025.00007 · Cited by 73
Sedláková, J., Lucivero, F., Pavarini, G., & Kerasidou, A. (2025) · The American Journal of Bioethics · doi:10.1080/15265161.2025.2526734 · Cited by 31
“Both aspects are problematic because of a technical limitation of CAI due to the limitation of datasets that do not represent all population groups (Bender et al. 2021; Weidinger et al. 2022).”
Han, C., Lu, X., & Fan, Q. (2025) · The Interpreter and Translator Trainer · doi:10.1080/1750399x.2025.2533606 · Cited by 29
“7 Since LLMs are inherently stochastic models that can sometimes generate random outputs based on the same input (Bender et al. 2021), we requested ChatGPT to perform automatic scoring five times for each source text, in parallel with the five human raters assigned to each interpreting task.”
Jukiewicz, M. (2025) · Frontiers in Education · doi:10.3389/feduc.2025.1594572 · Cited by 42
How Good Is NLP? A Sober Look at NLP Tasks through the Lens of Social Impact
Jin, Z., Chauhan, G., Tse, B., Sachan, M., & Mihalcea, R. F. (2021) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2021.findings-acl.273 · Cited by 44
“A current highly-debated topic in NLP ethics is GPT-3 (Brown et al., 2020), whose risks and harms include encoding gender and racist biases (Bender et al., 2021).”
BackgroundWhy the Data Revolution Needs Qualitative Thinking
Tanweer, A., Gade, E. K., Krafft, P. M., & Dreier, S. K. (2021) · Harvard Data Science Review · doi:10.1162/99608f92.eee0b0da · Cited by 56
Stammbach, D., Antoniak, M., & Ash, E. (2022) · doi:10.18653/v1/2022.wnu-1.6 · Cited by 74
Natural General Intelligence: How understanding the brain can help us build AI
Summerfield, C. (2022) · doi:10.1093/oso/9780192843883.001.0001 · Cited by 42
From Bias to Repair: Error as a Site of Collaboration and Negotiation in Applied Data Science Work
Lin, C. K., & Jackson, S. J. (2023) · Proceedings of the ACM on Human-Computer Interaction · doi:10.1145/3579607 · Cited by 22
“Natural Language Processing (NLP) experts Emily Bender et al. have similarly critiqued the rise of large language models, noting that many are built on internet datasets which skew toward younger populations, white men, and the Global North [12].”
BackgroundWhat do telecommunications policy academics have to fear from GPT-3?
Howell, B. E., & Potgieter, P. H. (2023) · Telecommunications Policy · doi:10.1016/j.telpol.2023.102576 · Cited by 22
Dirty Data Labeled Dirt Cheap: Epistemic Injustice in Machine Learning Systems
Hull, G. (2022) · Ethics and Information Technology · doi:10.1007/s10676-023-09712-y · Cited by 35
“Another is to acknowledge that the costs of ever larger training datasets may not be worth the gains in efficiency, and to favor a model of smaller, carefully curated datasets, worked out in conjunction with people who have relevant local knowledges (Bender et al., 2021).”
BackgroundWulff, P. (2024) · European Journal of Physics · doi:10.1088/1361-6404/ad0f9c · Cited by 43
Fagbohun, O., Harrison, R. M., & Dereventsov, A. (2023) · Journal of Artificial Intelligence Machine Learning and Data Science · doi:10.51219/jaimld/oluwole-fagbohun/15 · Cited by 70
“One central issue revolves around the potential for biased outputs 70,71 , reflecting discussions on algorithmic fairness and ethics.”
What Makes a Good Theory, and How Do We Make a Theory Good?
Guest, O. (2024) · Computational Brain & Behavior · doi:10.1007/s42113-023-00193-2 · Cited by 47
“…end of the spectrum, work such as that done in artificial intelligence (AI) with monstrously large artificial neural networks pollutes the atmosphere with climate-changing emissions at unprecedented scales (Strubell et al., 2019) for little benefit and arguably much harm (e.g. Bender et al., 2021).”
Tenório, K., & Romeike, R. (2023) · Proceedings of the 23rd Koli Calling International Conference on Computing Education Research · doi:10.1145/3631802.3631829 · Cited by 21
“…risks such as bias perpetuation, privacy and copyright violations, misinformation dissemination, manipulation facilitation, and security concerns [5, 43], researchers are also advocating the importance of these future professionals’ education about how to safely, critically, and ethically…”
Li, B., Wang, C., Bonk, C. J., & Kou, X. (2024) · TechTrends · doi:10.1007/s11528-024-00960-3 · Cited by 32
“It’s crucial to understand that, as critiqued in the “stochastic par-rot” view of Large Language Models (LLMs), these systems generate responses based on patterns in data without any inherent “wants” or “desires” (Bender et al., 2021).”
FAIR Enough: Develop and Assess a FAIR-Compliant Dataset for Large Language Model Training?
Raza, S., Ghuge, S., Ding, C., Dolatabadi, E., & Pandya, D. A. (2024) · Data Intelligence · doi:10.1162/dint_a_00255 · Cited by 49
“The importance of high-quality data cannot be overstated; it is highly important to use representative, well-curated datasets to develop models that are ethically sound and have broad applicability [7].”
BackgroundPalmer, Z. B., & Oswal, S. K. (2025) · Journal of Business and Technical Communication · doi:10.1177/10506519241280644 · Cited by 22
“As other studies have noted, we observed an extreme tendency toward generalized content creation and built-in biases in these website builders (Bender et al., 2021; Duin & Pedersen, 2023; Hutchinson et al., 2020).”
Mollema, W. J. T. (2025) · AI and Ethics · doi:10.1007/s43681-025-00801-w · Cited by 24
VeriPlan: Integrating Formal Verification and LLMs into End-User Planning
Lee, C. P., Porfirio, D., Wang, X. J., Zhao, K. C., & Mutlu, B. (2025) · Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3714113 · Cited by 21
“LLMs are prone to generating text that appears structurally coherent but contains factual inaccuracies or nonsensical information, a phenomenon known as hallucination [6, 32, 68, 83].”
Wang, J., Hu, H., Wang, Z., Yan, S., Sheng, Y., & He, D. (2024) · Proceedings of the CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3641917 · Cited by 21
Brunila, M., & LaViolette, J. (2021) · Joint SIGHUM Workshop on Computational Linguistics for Cultural Heritage, Social Sciences, Humanities and Literature (LaTeCH-CLfL) · doi:10.18653/v1/2021.latechclfl-1.18 · Cited by 21
“By reducing computational costs, GS pairing is additionally aligned with mounting calls for NLP and other ML researchers to intentfully pursue algorithms which minimize the growing environmental toll of their technologies (Bender et al., 2021; Strubell et al., 2019).”
BackgroundThe changing landscape of text mining: a review of approaches for ecology and evolution
Farrell, M. J., Guillarme, N. L., Brierley, L., Hunter, S. B., Scheepens, D. R., Willoughby, A. R., Yates, A., & Mideo, N. (2024) · Proceedings of the Royal Society B Biological Sciences · doi:10.1098/rspb.2024.0423 · Cited by 42
“These large language models (LLMs) push computational boundaries through a mix of new neural architectures and sheer increases in size (number of parameters, size of vocabulary, number of learned vectors, all reflecting a large amount of pre-training literature) [35,41,42].”
BackgroundDouglas, D. M., Lacey, J., & Howard, D. (2025) · AI and Ethics · doi:10.1007/s43681-024-00549-9 · Cited by 18
“Recognising the hardware necessary to develop and use an AI system ensures that the material characteristics of AI (such as the environmental impact and financial cost of operating it) are not overlooked [ 48].”
Integrating Content Moderation Systems with Large Language Models
Franco, M., Gaggi, O., & Palazzi, C. E. (2025) · ACM Transactions on the Web · doi:10.1145/3700789 · Cited by 38
System 2 Thinking in OpenAI’s o1-Preview Model: Near-Perfect Performance on a Mathematics Exam
de Winter, J., Dodou, D., & Eisma, Y. B. (2024) · Computers · doi:10.3390/computers13110278 · Cited by 32
“Indeed, while LLMs can produce fluent text, they struggle with seemingly basic tasks like counting [22], arithmetic [23,24], and reasoning [24,25].”
BackgroundMarino, E. B., Baleato, S., & Ribeiro, A. S. (2024) · Social Sciences · doi:10.3390/socsci13110603 · Cited by 54
AI-Augmented Predictions: LLM Assistants Improve Human Forecasting Accuracy
Schoenegger, P., Park, P. S., Karger, E., Trott, S., & Tetlock, P. E. (2025) · ACM Transactions on Interactive Intelligent Systems · doi:10.1145/3707649 · Cited by 69
“This is similar to work by Guo et al. [31] that provides a natural language interface for questions of tabular data.”
Simelane, P. M., & Kittur, J. (2024) · Computer Applications in Engineering Education · doi:10.1002/cae.22813 · Cited by 35
“Participants were particularly worried about students using the technology to bypass genuine learning, which could undermine the development of critical thinking and problem ‐ solving skills [69].”
Beninger, S., Reppel, A., Stanton, J., & Watson, F. (2025) · Journal of Marketing Education · doi:10.1177/02734753251316569 · Cited by 26
Agentic Workflows for Improving Large Language Model Reasoning in Robotic Object-Centered Planning
Moncada-Ramirez, J., Matez-Bandera, J.-L., González-Jiménez, J., & Ruiz-Sarmiento, J.-R. (2025) · Robotics · doi:10.3390/robotics14030024 · Cited by 35
Human agents, generative AI, and innovation: A formal model of hybrid creative process
Pedota, M., Cicala, F., & Basti, A. (2025) · Technovation · doi:10.1016/j.technovation.2025.103323 · Cited by 41
Jiao, J., Afroogh, S., Murali, A., Chen, K., Atkinson, D., & Dhurandhar, A. (2025) · Scientific Reports · doi:10.1038/s41598-025-18489-7 · Cited by 24
The Low-Resource Double Bind: An Empirical Study of Pruning for Low-Resource Machine Translation
Ahia, O., Kreutzer, J., & Hooker, S. (2021) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2021.findings-emnlp.282 · Cited by 74
“Over the years, the size of language models have grown exponentially (Amodei et al., 2018; Thompson et al., 2020; Bender et al., 2021).”
BackgroundUtilizing subjectivity level to mitigate identity term bias in toxic comments classification
Zhao, Z., Zhang, Z., & Hopfgartner, F. (2022) · Online Social Networks and Media · doi:10.1016/j.osnem.2022.100205 · Cited by 29
Human Evaluation of English–Irish Transformer-Based NMT
Lankford, S., Afli, H., & Way, A. (2022) · Information · doi:10.3390/info13070309 · Cited by 31
“Much of this effort has concentrated on NMT’s carbon footprint [47,48].”
BackgroundChatterjee, J., & Dethlefs, N. (2022) · IEEE Access · doi:10.1109/access.2022.3197167 · Cited by 28
The Doctor and the Algorithm: Promise, Peril, and the Future of Health AI
Graham, S. S. (2022) · doi:10.1093/oso/9780197644461.001.0001 · Cited by 32
Ethics and Machine Translation: The End User Perspective
Arenas, A. G., & Moorkens, J. (2023) · Machine translation · doi:10.1007/978-3-031-14689-3_7 · Cited by 46
CausalMapper: Challenging designers to think in systems with Causal Maps and Large Language Model
Huang, Z., Quan, K., Chan, J., & MacNeil, S. (2023) · Creativity and Cognition · doi:10.1145/3591196.3596818 · Cited by 24
The perils and promises of generative artificial intelligence in neurointerventional surgery
Ray, T. R., Kellogg, R. T., Fargen, K. M., Hui, F. K., & Vargas, J. (2023) · Journal of NeuroInterventional Surgery · doi:10.1136/jnis-2023-020353 · Cited by 26
Automating Empathy: Decoding Technologies that Gauge Intimate Life
McStay, A. (2023) · doi:10.1093/oso/9780197615546.001.0001 · Cited by 43
Finding Support Examples for In-Context Learning
Li, X., & Qiu, X. (2023) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2023.findings-emnlp.411 · Cited by 167
“• Language models have exhibited various kinds of bias (Bender et al., 2021), since our filtering stage is based on its feedback, the filtered example might also exhibit these biases.”
BackgroundEvaluating human resources management literacy: A performance analysis of ChatGPT and bard
Raman, R., Venugopalan, M., & Kamal, A. (2024) · Heliyon · doi:10.1016/j.heliyon.2024.e27026 · Cited by 48
“…termed " AI Hallucination, ” which refers to a scenario where AI-based systems generate inaccurate or nonsensical information Bender et al. [95], and our study has helped to validate the findings of this work and bolsters the view that a careful balance of utility and ethics have to…”
ResultThe potential of LLMs in hardware design
Alsaqer, S., Alajmi, S., Ahmad, I., & Alfailakawi, M. G. (2024) · Journal of Engineering Research · doi:10.1016/j.jer.2024.08.001 · Cited by 45
Teaching accounting in the era of ChatGPT – The student perspective
Sundkvist, C. H., & Kulset, E. M. (2024) · Journal of Accounting Education · doi:10.1016/j.jaccedu.2024.100932 · Cited by 81
Large Linguistic Models: Investigating LLMs’ Metalinguistic Abilities
Beguš, G., Dąbkowski, M. M., & Rhodes, R. E. (2025) · IEEE Transactions on Artificial Intelligence · doi:10.1109/tai.2025.3575745 · Cited by 78
“Proponents of symbolic processing models argue that neural networks lack systematicity, symbol grounding, meaning, abstract reasoning, and the ability to generalize beyond their training data [13], [14], [15], [36], [37].”
Automated Self-Supervised Learning for GraphsPreprint
Jin, W., Liu, X., Zhao, X., Ma, Y., Shah, N., & Tang, J. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2106.05470 · Cited by 91
“…works have illustrated the potential for representations learned in a self-supervised way to encode bias unintentionally, for example in language modeling (Bender et al., 2021; Zhao et al., 2019), image representation learning (Roberts et al., 2018) and outlier detection (Shekhar et al., 2020).”
Contrastive Language-Image Pre-training for the Italian Language
Bianchi, F., Attanasio, G., Pisoni, R., Terragni, S., Sarti, G., & Lakshmi, S. (2021) · University of Groningen research database (University of Groningen / Centre for Information Technology) · doi:10.48550/arxiv.2108.08688 · Cited by 54
Llama 2: Early Adopters' Utilization of Meta's New Open-Source Pretrained ModelPreprint
Roumeliotis, K. I., Tselikas, N. D., & Nasiopoulos, D. Κ. (2023) · Preprints.org · doi:10.20944/preprints202307.2142.v1 · Cited by 40
Good Models Borrow, Great Models Steal: Intellectual Property Rights and Generative AIPreprint
Chesterman, S. (2024) · Social Science Research Network · doi:10.2139/ssrn.4590006 · Cited by 118
Regulating under Uncertainty: Governance Options for Generative AIPreprint
G’Sell, F. (2024) · SSRN Electronic Journal · doi:10.2139/ssrn.4918704 · Cited by 42
Large Language Models in Critical Care Medicine: Scoping Review
Shi, T., Ma, J., Yu, Z., Xu, H., Yang, R., Xiong, M., Xiao, M., Li, Y., Zhao, H., & Kong, G. (2025) · JMIR Medical Informatics · doi:10.2196/76326 · Cited by 21
“Among them, hallucinations are one of the biggest drawbacks of LLMs, which make them act like stochastic parrots [62].”
Hallucinating with AI: Distributed Delusions and “AI Psychosis”
Osler, L. (2026) · Philosophy & Technology · doi:10.1007/s13347-026-01034-3 · Cited by 21
“Although many of us can spot something absurd in the recommendation to put glue on our pizza, when false outputs are pre-2 sented alongside accurate ones, it is easy for us to trust what generative AI places in front of us (Bender et al. 2021).”
There is no such thing as conscious artificial intelligence
Porębski, A., & Figura, J. (2025) · Humanities and Social Sciences Communications · doi:10.1057/s41599-025-05868-8 · Cited by 20
“Lacking intention and embeddedness in a communicative community, their language use is stripped of reference to meanings (those existing in the world rather than represented numerically in the vectors), making them comparable to stochastic parrots (Bender et al., 2021).”
How Do You Solve a Problem like DALL-E 2?
Wojtkiewicz, K. (2023) · Journal of Aesthetics and Art Criticism · doi:10.1093/jaac/kpad046 · Cited by 20
The Philosophy and Ethics of AI: Conceptual, Empirical, and Technological Investigations into Values
Simon, J., Rieder, G., & Branford, J. (2024) · Digital Society · doi:10.1007/s44206-024-00094-2 · Cited by 31
“One of these concerns the meanings of ‘meaning’ and ‘understanding’: While it seems clear that current LLMs do not understand but merely produce language patterns based upon probabilistic models, the question remains as to what else genuine understanding would require (Bender et al., 2021).”
Large language models and linguistic intentionality
Grindrod, J. (2024) · Synthese · doi:10.1007/s11229-024-04723-8 · Cited by 51
“Doing so will naturally require that we pay careful attention to the inner workings of LLMs.”
Osborne, C., Ding, J., & Kirk, H. R. (2024) · Journal of Computational Social Science · doi:10.1007/s42001-024-00300-8 · Cited by 94
“Following years of debate about the safety of openly releasing AI models [35, 18, 19, 36], recent years have seen the emergence and proliferation of “open” models, which individuals and organisations have shared on an open access basis on platforms such as the HF Hub [4].”
InfluentialBajohr, H. (2024) · Poetics Today · doi:10.1215/03335372-11092990 · Cited by 43
“…standard thus attained would, in turn, have an effect on human speakers again — it would have the status of a binding norm, integrated into all the mechanisms of writing that build on this technology, and which would be statistically almost impossible to escape (Bender et al. 2021; Bajohr 2024).”
BackgroundMuccione, V., Vaghefi, S. A., Bingler, J., Allen, S., Kraus, M., Gostlow, G., Wekhof, T., Senni, C. C., Stammbach, D., Ni, J., et al. (2024) · Regional Environmental Change · doi:10.1007/s10113-024-02283-8 · Cited by 28
“In absence of regular testing and evaluation, potentially well-crafted and convincing answers might hide incorrect or outdated statements (Bender et al. 2021).”
InfluentialAlshaigy, B., & Grande, V. (2024) · Nordic Conference on Human-Computer Interaction · doi:10.1145/3677045.3685493 · Cited by 34
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.