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 901–1,000 of 6,931
Human bias in AI models? Anchoring effects and mitigation strategies in large language models
Nguyen, J. K. (2024) · Journal of Behavioral and Experimental Finance · doi:10.1016/j.jbef.2024.100971 · Cited by 33
Neuberger, J., Ackermann, L., van der Aa, H., & Jablonski, S. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-75872-0_3 · Cited by 65
Dang, A., & Wang, H. (2024) · Journal of Second Language Writing · doi:10.1016/j.jslw.2024.101157 · Cited by 61
Competing narratives in AI ethics: a defense of sociotechnical pragmatism
Watson, D., Mökander, J., & Floridi, L. (2024) · AI & Society · doi:10.1007/s00146-024-02128-2 · Cited by 58
“Similarly, Bender et al. (2021) argue that the narrow utility of large language models must be balanced against a variety of costs—typi-cally born by those who do not benefit from the resulting technology—including financial burdens resulting in high barriers to market entry, as well as substantial…”
A review of university chatbots for student support: FAQs and beyond
Peyton, K., Unnikrishnan, S., & Mulligan, B. (2025) · Discover Education · doi:10.1007/s44217-025-00397-7 · Cited by 52
“[70] who suggest that these issues may be remedied by the careful curation of datasets relevant to the task at hand.”
A Taxonomy of Linguistic Expressions That Contribute To Anthropomorphism of Language Technologies
DeVrio, A., Cheng, M., Egede, L., Olteanu, A., & Blodgett, S. L. (2025) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3714038 · Cited by 85
Karjus, A. (2025) · Humanities and Social Sciences Communications · doi:10.1057/s41599-025-04503-w · Cited by 32
“These issues have been and will be discussed elsewhere (Bender et al. 2021; Binz and Schulz 2023; Rooij et al. 2024; Feng et al. 2023; Hagendorff et al. 2024; Liesenfeld et al. 2023; Lund et al. 2023; Motoki et al. 2023; Novozhilova et al. 2024; Ollion et al. 2024; Tomlinson et al. 2024).”
Developing generative AI literacies through self-regulated learning: A human-centered approach
Anders, A. D., & Speltz, E. D. (2025) · Computers and Education Artificial Intelligence · doi:10.1016/j.caeai.2025.100482 · Cited by 86
Algorithmic Factors Influencing Bias in Machine Learning
Blanzeisky, W., & Cunningham, P. (2021) · Communications in computer and information science · doi:10.1007/978-3-030-93736-2_41 · Cited by 49
“Despite the increasing number of related works over the past few years, the amount of anecdotal evidence of algorithmic bias is still growing [2].”
BackgroundLimitations of Autoregressive Models and Their Alternatives
Lin, C.-C., Jaech, A., Li, X., Gormley, M. R., & Eisner, J. (2021) · Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies · doi:10.18653/v1/2021.naacl-main.405 · Cited by 27
“This is downright depressing, considering the costs invested in training huge parametric autoregressive models (Bender et al., 2021): since ! ∈ P, it is trivial to build an efficient scoring function ?̃(x) with fixed parameters that has ! as its support, just not an autoregressive one.”
BackgroundActionable Approaches to Promote Ethical AI in Libraries
Bubinger, H., & Dinneen, J. D. (2021) · Proceedings of the Association for Information Science and Technology · doi:10.1002/pra2.528 · Cited by 45
“However, such systems must be developed, implemented, and maintained carefully to minimise the numerous ethical issues they entail, small and large, from data and design to deployment, including issues with bias, fairness, accountability, transparency, responsibility, environmental impact, and more (Bender et al., 202…”
BackgroundBiggiogera, J., Boateng, G., Hilpert, P., Vowels, M. J., Bodenmann, G., Neysari, M., Nußbeck, F. W., & Kowatsch, T. (2021) · Companion Publication of the International Conference on Multimodal Interaction (ICMI Companion) · doi:10.1145/3461615.3485423 · Cited by 45
On Natural Language User Profiles for Transparent and Scrutable Recommendation
Radlinski, F., Balog, K., Díaz, F., Dixon, L., & Wedin, B. (2022) · International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR) · doi:10.1145/3477495.3531873 · Cited by 70
On the Machine Learning of Ethical Judgments from Natural Language
Talat, Z., Blix, H., Valvoda, J., Ganesh, M. I., Cotterell, R., & Williams, A. (2022) · Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies · doi:10.18653/v1/2022.naacl-main.56 · Cited by 28
“However, recent critical scholarship on machine learning, and in particular on language models, argues that large-scale language models mimic without understanding (Bender et al., 2021), and don’t have communicative intent (Bender and Koller, 2020)— in short, they lack what is required.”
BackgroundScientists' Perspectives on the Potential for Generative AI in their Fields
Morris, M. R. (2023) · arXiv · doi:10.48550/arxiv.2304.01420 · Cited by 28
“…[9]), Generative AI still has many limitations, including potential risks such as hallucinations (fabrications, non-factual outputs, e.g. [18]), production of biased or offensive content, memorization of training data, and ethical concerns over the provenance of training materials [3, 54].”
BackgroundChurch, K. W., & Kordoni, V. (2022) · Natural Language Engineering · doi:10.1017/s1351324922000043 · Cited by 28
Dense and Disconnected: Analyzing the Sedimented Style of ChatGPT-Generated Text at Scale
Markey, B., Brown, D. W., Laudenbach, M., & Kohler, A. (2024) · Written Communication · doi:10.1177/07410883241263528 · Cited by 27
“While it is unclear exactly what texts comprise ChatGPT’s training data, Bender et al. (2021) point out that the model is only trained on 7% of English users.”
“Personhood and AI: Why large language models don’t understand us”
Browning, J. (2024) · AI & SOCIETY · doi:10.1007/s00146-023-01724-y · Cited by 27
“And LLMs still produce grossly insensitive or immoral responses regularly (Bender et al. 2021).”
Applying Human Values Theory to Software Engineering Practice: Lessons and Implications
Ferrario, M. A., & Winter, E. (2022) · IEEE Transactions on Software Engineering · doi:10.1109/tse.2022.3170087 · Cited by 37
“For example, Bender et al. urge significant budget allocation to data curation and documentation of training data for language models found to ‘encode biased views harmful to marginalized populations’ [51].”
BackgroundGender Bias in Masked Language Models for Multiple Languages
Kaneko, M., Imankulova, A., Bollegala, D., & Okazaki, N. (2022) · Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT) · doi:10.18653/v1/2022.naacl-main.197 · Cited by 92
Yokoyama, A., Ferro, M., de Paula, F. B., Vieira, V., & Schulze, B. R. (2023) · Concurrency and Computation Practice and Experience · doi:10.1002/cpe.7825 · Cited by 38
Dolmestch, R. N. (2023) · Derecho PUCP · doi:10.18800/derechopucp.202302.007 · Cited by 87
AI-Testimony, Conversational AIs and Our Anthropocentric Theory of Testimony
Freiman, O. (2024) · Social Epistemology · doi:10.1080/02691728.2024.2316622 · Cited by 43
“…and engineers by ex-Google engineer Blake Lemoine (Tiku 2022; cf.; Bryson 2022), for many philosophers it is clear that LLMs are not sentient, do not have minds, their lingual outputs are not the expression of thoughts but are rather only sophisticated statistical machines (Bender et al. 2021).”
BackgroundThe implications of handwritten text recognition for accessing the past at scale
Nockels, J., Gooding, P., & Terras, M. (2024) · Journal of Documentation · doi:10.1108/jd-09-2023-0183 · Cited by 47
Towards Environmentally Equitable AI via Geographical Load Balancing
Li, P., Yang, J., Wierman, A., & Ren, S. (2024) · ACM International Conference on Future Energy Systems (e-Energy) · doi:10.1145/3632775.3661938 · Cited by 60
Hype, Sustainability, and the Price of the Bigger-is-Better Paradigm in AI
Varoquaux, G. P., Luccioni, A. S., & Whittaker, M. (2025) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3715275.3732006 · Cited by 128
Posthuman Cartography? Rethinking Artificial Intelligence, Cartographic Practices, and Reflexivity
Lin, Y., & Zhao, B. (2025) · Annals of the American Association of Geographers · doi:10.1080/24694452.2024.2435920 · Cited by 46
“The use of larger and more complex models trained on vast amounts of indiscriminate Internet data is likely to exacerbate this issue, as these training sets, despite their size, often contain inherent representation bias (Bender et al. 2021).”
Is Ockham’s razor losing its edge? New perspectives on the principle of model parsimony
Dubova, M., Chandramouli, S., Gigerenzer, G., Grünwald, P., Holmes, W. R., Lombrozo, T., Marelli, M., Musslick, S., Nicenboim, B., Ross, L. N., et al. (2025) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2401230121 · Cited by 70
Lermann Henestrosa, A., & Kimmerle, J. (2025) · Computers in Human Behavior: Artificial Humans · doi:10.1016/j.chbah.2025.100142 · Cited by 26
Prompts Are Programs Too! Understanding How Developers Build Software Containing Prompts
Liang, J. T., Lin, M. M., Rao, N., & Myers, B. A. (2025) · Proceedings of the ACM on software engineering. · doi:10.1145/3729342 · Cited by 75
“Additionally, the stochastic nature of FMs [15] made it difficult for participants to predict the model’s behavior, as found in prior work [36]: “I have never been confident [in predicting an FM’s behavior]...and I don’t think I ever will” (P10).”
The Attention Economy and the Impact of Artificial Intelligence
Baeza-Yates, R. A., & Fayyad, U. M. (2021) · Springer eBooks · doi:10.1007/978-3-030-86144-5_18 · Cited by 37
“There are examples of gender bias (Caliskan et al. 2017), The Attention Economy and the Impact of Arti fi cial Intelligence race bias (Larson et al. 2016), religious bias (Abid et al. 2021), etc., and their impact has many rami fi cations (Bender et al. 2021).”
BackgroundMitigating Gender Bias in Distilled Language Models via Counterfactual Role Reversal
Gupta, U., Dhamala, J., Kumar, V., Verma, A., Pruksachatkun, Y., Krishna, S., Gupta, R., Chang, K.-W., Steeg, G. V., & Galstyan, A. (2022) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2022.findings-acl.55 · Cited by 60
Semantic Web technologies and bias in artificial intelligence: A systematic literature review
Lobo, P. R., Daga, E., Alani, H., & Fernández, M. (2022) · Semantic Web · doi:10.3233/sw-223041 · Cited by 49
NLPositionality: Characterizing Design Biases of Datasets and Models
Santy, S., Liang, J. T., Bras, R. J. L., Reinecke, K., & Sap, M. (2023) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2023.acl-long.505 · Cited by 168
Mustapha, K. B., Mustapha, K. B., Yap, E. H., & Abakr, Y. A. (2024) · Interactive Technology and Smart Education · doi:10.1108/itse-10-2023-0198 · Cited by 37
Preparing Educators and Students at Higher Education Institutions for an AI-Driven World
Magrill, J., & Magrill, B. (2024) · Teaching & Learning Inquiry The ISSOTL Journal · doi:10.20343/teachlearninqu.12.16 · Cited by 43
“The integration of GenAI in higher education has been a topic of interest for the past decade, beginning with the advent of advanced neural network models and accessible machine learning software (Bender et al. 2021).”
MethodConjuring algorithms: Understanding the tech industry as stage magicians
Nagy, P., & Neff, G. (2024) · New Media & Society · doi:10.1177/14614448241251789 · Cited by 26
“Bender et al. (2021) call such technologies stochastic parrots.”
Rethinking AI: bias in speech-recognition chatbots for ELT
Jeon, J., Lee, S., & Coronel-Molina, S. M. (2024) · ELT Journal · doi:10.1093/elt/ccae035 · Cited by 43
Know Thyself, Improve Thyself: Personalized LLMs for Self-Knowledge and Moral Enhancement
Giubilini, A., Mann, S. P., Voinea, C., Earp, B. D., & Savulescu, J. (2024) · Science and Engineering Ethics · doi:10.1007/s11948-024-00518-9 · Cited by 41
“Existing research confirms LLMs’ capabilities to adapt to user-specific political profiles (Simmons, 2023), encode moral preferences (Schrerrer et al., 2023), and to capture societal biases (Bender et al., 2021).”
The Escalating AI’s Energy Demands and the Imperative Need for Sustainable Solutions
León, M. (2024) · WSEAS TRANSACTIONS on SYSTEMS archive · doi:10.37394/23202.2024.23.46 · Cited by 35
A posthumanist approach to AI literacy
Wang, Z., & Wang, C. (2025) · Computers & composition/Computers and composition · doi:10.1016/j.compcom.2025.102933 · Cited by 28
Position is Power: System Prompts as a Mechanism of Bias in Large Language Models (LLMs)
Neumann, A., Kirsten, E., Zafar, M. B., & Singh, J. (2025) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3715275.3732038 · Cited by 78
Li, N., & Quan, S. J. (2023) · Environment and Planning B Urban Analytics and City Science · doi:10.1177/23998083231151688 · Cited by 26
“While data collecting and processing are not immune to subjectivity and bias (Bender et al., 2021), the data-driven approach nonetheless does not rely heavily on expert knowledge and experience in identifying urban form typologies.”
BackgroundA Sociotechnical Audit: Assessing Police Use of Facial Recognition
Radiya-Dixit, E., & Neff, G. (2023) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3593013.3594084 · Cited by 45
“As the FAccT community has shown, understanding algorithmic power requires understanding the power structures in which algorithmic systems operate [12, 18, 55, 58, 76, 81, 85, 98].”
BackgroundMollema, W. J. T. (2024) · Open Journal of Social Sciences · doi:10.4236/jss.2024.122032 · Cited by 26
“AI and ML models have been characterized warningly as “stochastic parrots” (Bender et al., 2021) and “refracturing mirror of the world” (Pasquinelli & Joler, 2021).”
The widespread adoption of large language model-assisted writing across society
Liang, W., Zhang, Y., Codreanu, M., Wang, J., Cao, H., & Zou, J. (2025) · Patterns · doi:10.1016/j.patter.2025.101366 · Cited by 35
Climate Change and Health: The next Challenge of Ethical AI
Fiske, A., Radhuber, I. M., Willem, T., Buyx, A., Celi, L. A., & McLennan, S. (2025) · The Lancet Global Health · doi:10.1016/s2214-109x(25)00124-x · Cited by 25
Simulacra as conscious exotica
Shanahan, M. (2026) · Inquiry · doi:10.1080/0020174x.2024.2434860 · Cited by 25
Formanek, M. (2024) · Journal of Librarianship and Information Science · doi:10.1177/09610006241241066 · Cited by 42
“The highlighted disparities in responses by language models closely align with the findings of current scholarly inquiries conducted by various authors (such as Abid et al., 2021; Bender et al., 2021; Chan, 2023; Li, 2023; Sajan and Kyle, 2023 among others).”
Satyanarayan, A., & Jones, G. M. (2024) · doi:10.21428/e4baedd9.2d7598a2 · Cited by 40
ChatGPT: towards AI subjectivity
D’Amato, K. (2024) · AI & Society · doi:10.1007/s00146-024-01898-z · Cited by 41
“Is ChatGPT merely a “stochastic parrot” (Bender et al. 2021) or a “Chinese room” (Searle 1980)?”
Optimising the use of electronic medical records for large scale research in psychiatry
Newby, D., Taylor, N., Joyce, D. W., & Winchester, L. M. (2024) · Translational Psychiatry · doi:10.1038/s41398-024-02911-1 · Cited by 39
Kraft, A., & Soulier, E. (2024) · The 2024 ACM Conference on Fairness, Accountability, and Transparency · doi:10.1145/3630106.3658981 · Cited by 25
Supermind Ideator: How Scaffolding Human-AI Collaboration Can Increase Creativity
Heyman, J. L., Rick, S., Giacomelli, G., Wen, H., Laubacher, R., Taubenslag, N., Knicker, M. S., Jeddi, Y., Ragupathy, P., Curhan, J. R., et al. (2024) · ACM Collective Intelligence Conference (CI) · doi:10.1145/3643562.3672611 · Cited by 58
“Two core facets underlie this concept: reflection and inspiration .”
Danish university policies on generative AI: Problems, assumptions and sustainability blind spots
Driessens, O., & Pischetola, M. (2024) · MedieKultur: Journal of media and communication research · doi:10.7146/mk.v40i76.143595 · Cited by 25
Eisenreich, A., Just, J., Gimenez-Jimenez, D., & Füller, J. (2024) · Technovation · doi:10.1016/j.technovation.2024.103123 · Cited by 54
AI chatbots: A disguised enemy for academic integrity?
Niloy, A. C., Hafiz, R., Hossain, B. M., Gulmeher, F., Sultana, N., Islam, K. F., Bushra, F., Islam, S., Hoque, S. I., Rahman, M., et al. (2024) · International Journal of Educational Research Open · doi:10.1016/j.ijedro.2024.100396 · Cited by 26
Large language models in neurosurgery: a systematic review and meta-analysis
Patil, A., Serrato, P., Chisvo, N. S., Arnaout, O. M., See, P. A., & Huang, K. T. (2024) · Acta Neurochirurgica · doi:10.1007/s00701-024-06372-9 · Cited by 32
Fostering effective hybrid human-LLM reasoning and decision making
Passerini, A., Gema, A. P., Minervini, P., Sayin, B., & Tentori, K. (2025) · Frontiers in Artificial Intelligence · doi:10.3389/frai.2024.1464690 · Cited by 52
Intersectional analysis for science and technology
Nielsen, M. W., Gissi, E., Heidari, S., Horton, R. C., Nadeau, K. C., Ngila, D., Noble, S., Paik, H. Y., Tadesse, G. A., Zeng, E. Y., et al. (2025) · Nature · doi:10.1038/s41586-025-08774-w · Cited by 43
“When models designed for general applications are exposed to a wide range of images, linguistic styles, topics and cultural contexts, they perform more accurately across various scenarios 125,126 .”
Inie, N., Falk, J., & Selvan, R. (2025) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3714227 · Cited by 43
Emerging AI individualism: how young people integrate social AI into everyday life
Brandtzaeg, P. B., Følstad, A., & Skjuve, M. (2025) · Communication and Change · doi:10.1007/s44382-025-00011-2 · Cited by 30
Accelerating Deep Learning Inference: A Comparative Analysis of Modern Acceleration Frameworks
Ratul, I. J., Zhou, Y., & Yang, K. (2025) · Electronics · doi:10.3390/electronics14152977 · Cited by 49
On the Gap between Adoption and Understanding in NLP
Bianchi, F., & Hovy, D. (2021) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2021.findings-acl.340 · Cited by 35
“Parameter sizes have grown exponentially for the last few years (Bender et al., 2021), and what was considered preposterously large five years ago is now pretty standard.”
BackgroundBeyond Preserved Accuracy: Evaluating Loyalty and Robustness of BERT Compression
Xu, C., Zhou, W., Ge, T., Xu, K., McAuley, J., & Wei, F. (2021) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2021.emnlp-main.832 · Cited by 54
Pre-train or Annotate? Domain Adaptation with a Constrained Budget
Bai, F., Ritter, A., & Xu, W. H. (2021) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2021.emnlp-main.409 · Cited by 43
“However, modern pre-training methods incur substantial costs (Izsak et al., 2021), and generate carbon emissions (Strubell et al., 2019; Schwartz et al., 2020; Bender et al., 2021).”
BackgroundResponsible Language Technologies: Foreseeing and Mitigating Harms
Blodgett, S. L., Liao, Q. V., Olteanu, A., Mihalcea, R. F., Muller, M. J., Scheuerman, M. K., Tan, C., & Yang, Q. (2022) · CHI Conference Extended Abstracts on Human Factors in Computing Systems (CHI EA) · doi:10.1145/3491101.3516502 · Cited by 47
“Harmful impacts from language technologies can take many forms, from disparity in quality-of-service and unequal distribution of resources, to erasure, stereotyping and misrepresentation of groups and individuals [4, 5, 18].”
BackgroundNeural Pipeline for Zero-Shot Data-to-Text Generation
Kasner, Z., & Dušek, O. (2022) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2022.acl-long.271 · Cited by 50
DrBERT: A Robust Pre-trained Model in French for Biomedical and Clinical domains
Labrak, Y., Bazoge, A., Dufour, R., Rouvier, M., Morin, E., Daille, B., & Gourraud, P. (2023) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2023.acl-long.896 · Cited by 74
Centering health equity in large language model deployment
Singh, N., Lawrence, K., Richardson, S., & Mann, D. M. (2023) · PLOS Digital Health · doi:10.1371/journal.pdig.0000367 · Cited by 25
Software Testing of Generative AI Systems: Challenges and Opportunities
Aleti, A. (2023) · IEEE/ACM International Conference on Software Engineering: Future of Software Engineering (ICSE-FoSE) · doi:10.1109/icse-fose59343.2023.00009 · Cited by 72
Foundation Models in Healthcare: Opportunities, Risks & Strategies Forward
Thieme, A., Nori, A., Ghassemi, M., Bommasani, R., Andersen, T. O., & Luger, E. (2023) · Extended Abstracts of the 2023 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3544549.3583177 · Cited by 25
“…use within sensitive, high-stakes domains like healthcare to not accentuate risks of harm (i.e., how the use of FM-based applications may exacerbate social inequalities) [5, 39]; and raises fundamental questions about the responsible, ethical and safe use of such technologies going forward [2].”
InfluentialSystemic Biases in Sign Language AI Research: A Deaf-Led Call to Reevaluate Research Agendas
Desai, A., De Meulder, M., Hochgesang, J. A., Kocab, A., & Lu, A. X. (2024) · Proceedings of the Language Resources and Evaluation Conference · doi:10.63317/4qt6t64rxwd2 · Cited by 25
Ncube, P. D. N., Dzvapatsva, G. P., Matobobo, C., & Ranga, M. M. (2026) · AI and Ethics · doi:10.1007/s43681-025-00871-w · Cited by 31
“If not carefully designed, AI grading systems risk perpetuating biases present in training datasets [ 45, 46].”
The Role of Inclusion, Control, and Ownership in Workplace AI-Mediated Communication
Kadoma, K., Aubin Le Quere, M., Fu, X. J., Munsch, C., Metaxa, D., & Naaman, M. (2024) · Proceedings of the CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642650 · Cited by 24
“…tax” for individuals who do not align with these norms [39], affecting their perceived inclusion, control, and ownership over their work output. e However, as others have noted, one potential challenge is that language models can reflect social biases in the language they produce [4, 39].”
BackgroundSwiatek, L. M., Galloway, C., Vujnović, M., et al. (2023) · Public Relations Inquiry · doi:10.1177/2046147x231221828 · Cited by 24
Postdigital Duoethnography: An Inquiry into Human-Artificial Intelligence Synergies
Brailas, A. (2024) · Postdigital Science and Education · doi:10.1007/s42438-024-00455-7 · Cited by 37
“LLMs and ChatGPT in particular can successfully serve this role, even if we think of them as mere ‘stochastic parrots’ (Bender et al. 2021).”
The Promises and Perils of Foundation Models in Dermatology
Gui, H., Omiye, J. A., Chang, C. T., & Daneshjou, R. (2024) · Journal of Investigative Dermatology · doi:10.1016/j.jid.2023.12.019 · Cited by 30
Systematic analysis of 32,111 AI model cards characterizes documentation practice in AI
Liang, W., Rajani, N. F., Yang, X., Ozoani, E., Wu, E., Chen, Y. T., Smith, D. S., & Zou, J. (2024) · Nature Machine Intelligence · doi:10.1038/s42256-024-00857-z · Cited by 101
Artificial Dreams: Surreal Visual Storytelling as Inquiry Into AI 'Hallucination'
Halperin, B. A., & Lukin, S. M. (2024) · Designing Interactive Systems Conference · doi:10.1145/3643834.3660685 · Cited by 35
“Amid the rise of generative artifcial intelligence (AI) media, there is a ballooning amount of concern around the generation of false, deceptive, and hateful narratives as well as representations through multiple modalities, including text [12], image [13], and video [44] with sound [34].”
BackgroundInfluentialde Roock, R. S. (2024) · Reading Research Quarterly · doi:10.1002/rrq.569 · Cited by 47
Generative AI and children’s digital futures: New research challenges
Leaver, T. B., & Srdarov, S. (2025) · Journal of Children and Media · doi:10.1080/17482798.2024.2438679 · Cited by 25
“Instead, this is the work of extremely sophisticated probabilistic models (Bender et al., 2021).”
Reinforcement Learning from Human Feedback in LLMs: Whose Culture, Whose Values, Whose Perspectives?
Barman, K. G., Lohse, S., & de Regt, H. W. (2025) · Philosophy & Technology · doi:10.1007/s13347-025-00861-0 · Cited by 58
“They have even been described as “stochastic parrots” (Bender et al., 2021) and “bullshit genera-tors” (Hicks et al., 2024).”
Simulacrum of Stories: Examining Large Language Models as Qualitative Research Participants
Kapania, S., Agnew, W. S., Eslami, M., Heidari, H., & Fox, S. (2025) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3713220 · Cited by 86
“LMs fundamentally work by detecting the statistical patterns present in natural language data [13].”
Kozlowski, A. C., & Evans, J. A. (2025) · Sociological Methods & Research · doi:10.1177/00491241251337316 · Cited by 72
Toward Sustainability in 6G and Beyond: Challenges and Opportunities of Open RAN
Ahmadi, H., Rahmani, M., Chetty, S. B., Tsiropoulou, E. E., Arslan, H., Debbah, M. A., & Quek, T. Q. S. (2025) · IEEE Communications Standards Magazine · doi:10.1109/mcomstd.2025.3575488 · Cited by 25
“This amount of consumed energy also leads to a huge CO 2 emissions; as presented in [11] training a big transformer model causes more CO 2 emission than 55 average person per year.”
Practices, opportunities and challenges in the fusion of knowledge graphs and large language models
Cai, L., Yu, C., Kang, Y., Fu, Y., Zhang, H., & Zhao, Y. (2025) · Frontiers in Computer Science · doi:10.3389/fcomp.2025.1590632 · Cited by 48
ALGORITHMIC BIAS IN AI-ENHANCED EDUCATION: CULTURAL DIMENSIONS AND PEDAGOGICAL IMPACT
Fadil, H. O. C. A., & Nuredin, A. (2025) · Symposium Proceedings · doi:10.55843/isl2025symp163h · Cited by 27
EnergyVis: Interactively Tracking and Exploring Energy Consumption for ML Models
Shaikh, O., Saad-Falcon, J., Wright, A. P., Das, N., Freitas, S., Asensio, O. I., & Chau, D. H. (2021) · CHI Conference Extended Abstracts on Human Factors in Computing Systems (CHI EA) · doi:10.1145/3411763.3451780 · Cited by 39
Wiese, W., & Friston, K. J. (2021) · Behavioural Brain Research · doi:10.1016/j.bbr.2021.113704 · Cited by 27
“Examples include applications with racist or other biases [73].”
BackgroundBaeza-Yates, R. A. (2022) · ACM International Conference on Web Search and Data Mining (WSDM) · doi:10.1145/3488560.3498370 · Cited by 38
“The best example are large language models [5],”
BackgroundRepresenting Context in FrameNet: A Multidimensional, Multimodal Approach
Torrent, T. T., Matos, E. E. D. S., Belcavello, F., Viridiano, M., Gamonal, M. A., da Costa, A. D., & Marim, M. C. (2022) · Frontiers in Psychology · doi:10.3389/fpsyg.2022.838441 · Cited by 46
“…and ELMo (Peters et al., 2018) have brought contextual information to the core of Natural Language Processing (NLP), raising key issues on which kinds of information are captured by and on how they are represented in those models (Rogers et al., 2020; Bender et al., 2021; Bommasani et al., 2021).”
BackgroundEfficient Hierarchical Domain Adaptation for Pretrained Language Models
Chronopoulou, A., Peters, M., & Dodge, J. (2022) · Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies · doi:10.18653/v1/2022.naacl-main.96 · Cited by 24
“As such models are trained on large datasets from text in the Internet, they encode biases that could harm marginalized populations (Bender et al., 2021).”
BackgroundSchäfer, M. S., & Hase, V. (2022) · Wiley Interdisciplinary Reviews Climate Change · doi:10.1002/wcc.806 · Cited by 30
“Researchers should consider whether they want to contribute to the affectedness of populations not benefitting from such models by building NLP approaches that “ serve the needs of those who already have the most privilege in society ” (Bender et al., 2021, p. 613).”
BackgroundRocca, R., Tamagnone, N., Fekih, S., Contla, X., & Rekabsaz, N. (2023) · Frontiers in Big Data · doi:10.3389/fdata.2023.1082787 · Cited by 47
“Thirdly, it is widely known that publicly available NLP models can absorb and reproduce multiple forms of biases (e.g., racial or gender biases Bolukbasi et al., 2016; Davidson et al., 2019; Bender et al., 2021).”
BackgroundArtificial intelligence for global health: cautious optimism with safeguards
Pujari, S., Reis, A. A., Zhao, Y., Alsalamah, S., Serhan, F., Reeder, J. C., & Labrique, A. B. (2023) · Bulletin of the World Health Organization · doi:10.2471/blt.23.290215 · Cited by 31
An Empirical Analysis of Parameter-Efficient Methods for Debiasing Pre-Trained Language Models
Xie, Z., & Lukasiewicz, T. (2023) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2023.acl-long.876 · Cited by 58
Mutual Information Alleviates Hallucinations in Abstractive Summarization
van der Poel, L., Cotterell, R., & Meister, C. (2022) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2022.emnlp-main.399 · Cited by 105
Wieser, M., Schoeffmann, K., Stefanics, D., Bollin, A., & Pasterk, S. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-44900-0_4 · Cited by 34
Design of a large language model for improving customer service in telecom operators
Xiaoliang, M., RuQiang, Z., Ying, L., Deng, C., & Dequan, D. (2024) · Electronics Letters · doi:10.1049/ell2.13218 · Cited by 33
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.