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 501–600 of 6,931
Ferrario, A., Sedlakova, J., & Trachsel, M. (2024) · JMIR Mental Health · doi:10.2196/56569 · Cited by 61
“In contrast, LLMs lack these commitments [40].”
Chakrabarty, T., Padmakumar, V., Brahman, F., & Muresan, S. (2024) · Creativity and Cognition · doi:10.1145/3635636.3656201 · Cited by 61
The BigScience ROOTS Corpus: A 1.6TB Composite Multilingual Dataset
Laurençon, H., Saulnier, L., Wang, T., Akiki, C., del Moral, A. V., Scao, T. L., Von Werra, L., Mou, C., Ponferrada, E. G., Nguyen, H., et al. (2023) · arXiv · doi:10.48550/arxiv.2303.03915 · Cited by 93
Self-citationSurvey on Factuality in Large Language Models: Knowledge, Retrieval and Domain-Specificity
Wang, C., Liu, X., Yue, Y., Tang, X., Zhang, T., Jiayang, C., Yao, Y., Gao, W., Hu, X., Qi, Z., et al. (2023) · arXiv · doi:10.48550/arxiv.2310.07521 · Cited by 61
Truth machines: synthesizing veracity in AI language models
Munn, L., Magee, L., & Arora, V. (2024) · AI & SOCIETY · doi:10.1007/s00146-023-01756-4 · Cited by 65
DEMix Layers: Disentangling Domains for Modular Language Modeling
Gururangan, S., Lewis, M., Holtzman, A., Smith, N., & Zettlemoyer, L. (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.407 · Cited by 60
Schramowski, P., Tauchmann, C., & Kersting, K. (2022) · 2022 ACM Conference on Fairness Accountability and Transparency · doi:10.1145/3531146.3533192 · Cited by 63
“The entailed issues have been discussed for language models, for instance, models producing stereotypical and derogatory content [Bender et al. 2021], and for vision model respectively CV datasets highlighting, e.g., gender and racial biases [Denton et al. 2021; Larrazabal et al. 2020; Steed and…”
BackgroundYou reap what you sow: On the Challenges of Bias Evaluation Under Multilingual Settings
Talat, Z., Névéol, A., Biderman, S., Clinciu, M., Dey, M., Longpre, S., Luccioni, S., Masoud, M., Mitchell, M., Radev, D., et al. (2022) · Proceedings of BigScience Episode #5 -- Workshop on Challenges & Perspectives in Creating Large Language Models · doi:10.18653/v1/2022.bigscience-1.3 · Cited by 59
Fleming, G. C., Klopfer, M. D., Katz, A., & Knight, D. B. (2024) · Journal of Engineering Education · doi:10.1002/jee.20581 · Cited by 94
“…this choice because the scale of the data collected and relatively standardized format meant that we could predictably capture information that provides insight into this phenomenon of skill demand without using models critiqued for their black box characteristics (Bender et al., 2021; Li, 2018).”
BackgroundUnsettling Languages, Unruly Bodyminds: Imaging a Crip LinguisticsPreprint
Henner, J., & Robinson, O. E. (2021) · PsyArXiv (OSF Preprints) · doi:10.31234/osf.io/7bzaw · Cited by 68
Experimental narratives: A comparison of human crowdsourced storytelling and AI storytelling
Beguš, N. (2024) · Humanities and Social Sciences Communications · doi:10.1057/s41599-024-03868-8 · Cited by 60
“LLMs mimic patterns found in their training data, which may lead to the ampli fi cation of social biases (Bender et al. 2021).”
Hocky, G. M., & White, A. D. (2022) · Digital Discovery · doi:10.1039/d1dd00009h · Cited by 59
“It can be unsettling too because it has quite adeptly captured the racism, sexism, and bias in human writing and can be reflected in its responses [15].”
Arpacı, İ. (2023) · IEEE Transactions on Engineering Management · doi:10.1109/tem.2023.3339578 · Cited by 73
“In addition, AI chatbots may generate biased or inaccurate content due to biases in the training data or underlying algorithms [15].”
BackgroundKarpinska, M., & Iyyer, M. (2023) · Proceedings of the Eighth Conference on Machine Translation · doi:10.18653/v1/2023.wmt-1.41 · Cited by 65
Exploring the frontiers of LLMs in psychological applications: a comprehensive review
Ke, L., Tong, S., Cheng, P., & Peng, K. (2025) · Artificial Intelligence Review · doi:10.1007/s10462-025-11297-5 · Cited by 149
“Causal Reasoning and Creativity Sartori and Orrù (2023) Poor causal reasoning, dependence on biased training data, lack of creativity and imagination.”
To protect science, we must use LLMs as zero-shot translators
Mittelstadt, B., Wachter, S., & Russell, C. (2023) · Nature Human Behaviour · doi:10.1038/s41562-023-01744-0 · Cited by 58
“Downsizing the size and scope of models should also be considered, including restricting the types of valid prompts to areas for which the model has reliable training data or can pass an accepted confidence threshold 1 .”
BackgroundBlanchard, A., & Taddeo, M. (2023) · Digital Society · doi:10.1007/s44206-023-00036-4 · Cited by 79
“See Bender (2022), Bender and Koller (2020), and Bender et al. (2021) on the limitations of large language models.”
BackgroundOn reading and interpreting black box deep neural networks
Dobson, J. E. (2023) · International Journal of Digital Humanities · doi:10.1007/s42803-023-00075-w · Cited by 54
“(Bender et al., 2021).”
Artificial Intelligence and Sustainability—A Review
Dhiman, R., Miteff, S., Wang, Y., Ma, S.-C., Amirikas, R., & Fabian, B. (2024) · Analytics · doi:10.3390/analytics3010008 · Cited by 59
Medical Ethics of Large Language Models in Medicine
Ong, J. C. L., Chang, S. Y.-H., William, W., Butte, A. J., Shah, N. H., Chew, L. S. T., Liu, N., Doshi-Velez, F., Lu, W., Savulescu, J., et al. (2024) · NEJM AI · doi:10.1056/aira2400038 · Cited by 58
Symeou, L., Louca, L., Kavadella, A., Mackay, J., Danidou, Y., & Raffay, V. (2025) · European Journal of Dental Education · doi:10.1111/eje.13069 · Cited by 60
Gender biases within Artificial Intelligence and ChatGPT: Evidence, Sources of Biases and Solutions
Ho, J. Q. H., Hartanto, A., Koh, A., & Majeed, N. M. (2025) · Computers in Human Behavior Artificial Humans · doi:10.1016/j.chbah.2025.100145 · Cited by 112
Bergier, H., Duron, L., Sordet, C., Kawka, L., Schlencker, A., Chasset, F., & Arnaud, L. (2021) · Autoimmunity Reviews · doi:10.1016/j.autrev.2021.102864 · Cited by 69
Putting Psychology to the Test: Rethinking Model Evaluation Through Benchmarking and Prediction
Rocca, R., & Yarkoni, T. (2021) · Advances in Methods and Practices in Psychological Science · doi:10.1177/25152459211026864 · Cited by 63
“Recent literature has highlighted a number of scientific and ethical shortcomings associated with common machine-learning practices, especially with respect to recent developments in natural language processing (for a comprehensive overview, see Bender et al., 2021; Bender & Koller, 2020).”
BackgroundTREC Deep Learning Track: Reusable Test Collections in the Large Data Regime
Craswell, N., Mitra, B., Yilmaz, E., Campos, D., Voorhees, E. M., & Soboroff, I. (2021) · Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval · doi:10.1145/3404835.3463249 · Cited by 58
The Illusion of Artificial Inclusion
Agnew, W., Bergman, A. S., Chien, J., Díaz, M., El-Sayed, S., Pittman, J., Mohamed, S., & McKee, K. R. (2024) · Proceedings of the CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642703 · Cited by 58
“Critics argue that such claims of understanding and general intelligence generate overconfidence in model abilities, which in turn could deter efforts at oversight and encourage unsafe or inappropriate applications [19, 126, 145].”
BackgroundGenerative AI as source of change of knowledge management paradigm
Kaczorowska–Spychalska, D., Kotula, N., Mazurek, G., & Sułkowski, Ł. (2024) · Human Technology · doi:10.14254/1795-6889.2024.20-1.7 · Cited by 100
“It is also good at producing content that lacks any legitimate basis (Mollick & Mollick, 2022) or is laden with human biases, which increases its level of bias (Bender et al., 2021; Haluza & Jungwirth, 2023).”
BackgroundEmerging Roles and Relationships Among Humans and Interactive AI Systems
Lindgren, H. (2024) · International Journal of Human-Computer Interaction · doi:10.1080/10447318.2024.2435693 · Cited by 102
“Studies are currently exploring how large language models (LLM) could be integrated in dialogues to achieve purposeful activities overcoming their limitations, in particular, reliability (correctness of the information) (Bender et al., 2021).”
Cambria, E., Zhang, X., Mao, R., Chen, M., & Kwok, K. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-76827-9_11 · Cited by 97
“Some researchers defined large language models (LLMs) like ChatGPT as ‘stochastic parrots’ [4], i.e., systems that haphazardly stitch together sequences of linguistic forms it has observed in its vast training data, according to probabilistic information about how they combine, but without any…”
BackgroundSystematic literature review on bias mitigation in generative AI
Afreen, J., Mohaghegh, M., & Doborjeh, M. (2025) · AI and Ethics · doi:10.1007/s43681-025-00721-9 · Cited by 92
Generative AI and Jobs: A Global Analysis of Potential Effects on Job Quantity and QualityPreprint
Gmyrek, P., Berg, J., & Bescond, D. (2023) · SSRN Electronic Journal · doi:10.2139/ssrn.4584219 · Cited by 134
Delobelle, P., Tokpo, E., Calders, T., & Berendt, B. (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.122 · Cited by 57
“We also did not draw much attention to many other negative impacts of language models that practitioners should consider, e.g. high energy usage or not including all stakeholders when training a language model (Bender et al., 2021).”
BackgroundVisText: A Benchmark for Semantically Rich Chart Captioning
Tang, B. J., Boggust, A., & Satyanarayan, A. (2023) · arXiv · doi:10.48550/arxiv.2307.05356 · Cited by 57
“To what degree are chart captioning models stochastic parrots (Bender et al., 2021), and how much do they understand the information charts depict?”
BackgroundInsolvent: How to Reorient Computing for Just Sustainability
Becker, C. (2023) · doi:10.7551/mitpress/14668.001.0001 · Cited by 56
Ecovisor: A Virtual Energy System for Carbon-Efficient Applications
Souza, A., Bashir, N., Murillo, J., Hanafy, W. A., Liang, Q., Irwin, D., & Shenoy, P. J. (2023) · International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS) · doi:10.1145/3575693.3575709 · Cited by 123
“Recent work has recognized the importance of reducing carbon emissions, and has attempted to quantify the carbon emissions of running particular applications [25, 29, 57, 66].”
BackgroundSon, J., & Kim, B. (2023) · Information · doi:10.3390/info14100574 · Cited by 100
Characterizing Manipulation from AI Systems
Carroll, M., Chan, A., Ashton, H., & Krueger, D. (2023) · Equity and Access in Algorithms, Mechanisms, and Optimization · doi:10.1145/3617694.3623226 · Cited by 60
“Langosco et al. [102] and Shah et al. [146] show that both language models and general RL agents can pursue different goals in out-of-distribution environments even when trained to perfect accuracy on in-distribution environments.”
BackgroundFairness in Deep Learning: A Survey on Vision and Language Research
Parraga, O., Móre, M. D., Mattjie, C., Gavenski, N. S., Kupssinskü, L. S., Medronha, A., de Moura, L. V., Simões, G. S., & Barros, R. C. (2025) · ACM Computing Surveys · doi:10.1145/3637549 · Cited by 125
A multimodal grammar of artificial intelligence: Measuring the gains and losses in generative AI
Cope, B., & Kalantzis, M. (2023) · Multimodality & Society · doi:10.1177/26349795231221699 · Cited by 86
“Some have argued that Generative AI is a “ stochastic parrot ” of sorts, as if it just repeats textual formulae that it has been taught (Bender et al., 2021).”
BackgroundKim, Y., Lee, J., Kim, S., Park, J., & Kim, J. (2024) · Proceedings of the 29th International Conference on Intelligent User Interfaces · doi:10.1145/3640543.3645148 · Cited by 55
Open challenges and opportunities in federated foundation models towards biomedical healthcare
Li, X., Peng, L., Wang, Y., & Zhang, W. (2025) · BioData Mining · doi:10.1186/s13040-024-00414-9 · Cited by 108
MS MARCO: Benchmarking Ranking Models in the Large-Data Regime
Craswell, N., Mitra, B., Yilmaz, E., Campos, D., & Lin, J. (2021) · Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval · doi:10.1145/3404835.3462804 · Cited by 55
Linguistically inspired roadmap for building biologically reliable protein language models
Vu, M. H., Akbar, R., Robert, P. A., Swiatczak, B., Sandve, G. K., Greiff, V., & Haug, D. T. T. (2023) · Nature Machine Intelligence · doi:10.1038/s42256-023-00637-1 · Cited by 54
Schetinger, V., Bartolomeo, S. D., El‐Assady, M., McNutt, A. M., Miller, M., Passos, J. P. A., & Adams, J. L. (2023) · Computer Graphics Forum · doi:10.1111/cgf.14841 · Cited by 84
Science communication with generative AI
Alvarez, A., Caliskan, A., Crockett, M. J., Ho, S. S., Messeri, L. R., & West, J. D. (2024) · Nature Human Behaviour · doi:10.1038/s41562-024-01846-3 · Cited by 81
GraphTranslator: Aligning Graph Model to Large Language Model for Open-ended Tasks
Zhang, M., Sun, M., Wang, P., Fan, S., Mo, Y., Xu, X., Liu, H., Yang, C. H., & Shi, C. (2024) · International World Wide Web Conference (WWW) · doi:10.1145/3589334.3645682 · Cited by 146
“In future work, we can use larger-scale LLMs such as ChatGPT, to improve the quality of the generated text descriptions.”
MethodEvaluating literature reviews conducted by humans versus ChatGPT: comparative study
Mostafapour, M., Fortier, J. H., Pacheco, K., Murray, H., & Garber, G. (2024) · JMIR AI · doi:10.2196/56537 · Cited by 55
Testing AI on language comprehension tasks reveals insensitivity to underlying meaning
Dentella, V., Günther, F., Murphy, E., Marcus, G., & Leivada, E. (2024) · Scientific Reports · doi:10.1038/s41598-024-79531-8 · Cited by 54
“Limited information: While the sentence “John read to himself, Anthony and Franck” implies John read aloud to others, it doesn’t clarify if Franck was one of the recipients or just present during the reading.”
Forgetting Practices in the Data Sciences
Muller, M., & Strohmayer, A. (2022) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3491102.3517644 · Cited by 55
Harm to Nonhuman Animals from AI: a Systematic Account and Framework
Coghlan, S., & Parker, C. (2023) · Philosophy & Technology · doi:10.1007/s13347-023-00627-6 · Cited by 77
Wörsdörfer, M. (2023) · Global Business and Organizational Excellence · doi:10.1002/joe.22238 · Cited by 80
Farina, M., Yu, X., & Lavazza, A. (2024) · AI and Ethics · doi:10.1007/s43681-023-00405-2 · Cited by 75
“At issue here is not whether LLMs are intelligent (sentient) or whether instead they should be considered as ‘stochastic parrots’ [5].”
The Limitations of Large Language Models for Understanding Human Language and Cognition
Cuskley, C. F., Woods, R. J. H., & Flaherty, M. (2024) · Open Mind · doi:10.1162/opmi_a_00160 · Cited by 164
“This isn ’ t “ merely ” an issue of ethics and inclusion in scientific practice (a fundamental problem for AI that extends well beyond this; Bender et al., 2021; Birhane & van Dijk, 2020; Erscoi et al., 2023; Rillig et al., 2023), but risks leading us to only attempting to understand a confined…”
BackgroundMERLOT: Multimodal Neural Script Knowledge Models
Zellers, R., Lu, X., Hessel, J., Yu, Y., Park, J. S., Cao, J., Farhadi, A., & Choi, Y. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2106.02636 · Cited by 53
“Additionally, there are inherent social biases on YouTube – and treating these videos as equivalent to ‘the world’ [111] can embed hegemonic perspectives [42, 114, 13].”
BackgroundStory Centaur: Large Language Model Few Shot Learning as a Creative Writing Tool
Swanson, B., Mathewson, K., Pietrzak, B., Chen, S., & Dinalescu, M. (2021) · Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: System Demonstrations · doi:10.18653/v1/2021.eacl-demos.29 · Cited by 53
“One of the most promising possibilities for large language models (LLMs) is few-shot learning (Brown et al., 2020) in which it is possible to create text to text models with little or no training data.”
MethodInfluentialThe seer and the seen: Surveying Palantir’s surveillance platform
Iliadis, A., & Acker, A. (2022) · The Information Society · doi:10.1080/01972243.2022.2100851 · Cited by 82
An elemental ethics for artificial intelligence: water as resistance within AI’s value chain
Lehuedé, S. (2024) · AI & Society · doi:10.1007/s00146-024-01922-2 · Cited by 104
“…represent a quite limited set of impacts that, by focusing on issues that can be technically fixed, can leave unchecked relevant structural injustices underpinning the AI industry (Attard-Frost and Widder 2023; Ricaurte 2022; Bender et al. 2021; Powell et al. 2022; Rességuier and Rodrigues 2020).”
Making It Possible for the Auditing of AI: A Systematic Review of AI Audits and AI Auditability
Li, Y., & Goel, S. (2024) · Information Systems Frontiers · doi:10.1007/s10796-024-10508-8 · Cited by 107
“A few large models like ChatGPT are trained with large amounts of data that privilege quantity over quality of content, with little attention paid to the social context, thereby making the audit more difficult (Bender et al., 2021; Berghoff et al., 2021).”
The Dawn of the Human-Machine Era: A forecast of new and emerging language technologies
Sayers, D., Sousa‐Silva, R., Höhn, S., Ahmedi, L., Allkivi-Metsoja, K., Anastasiou, D., Beňuš, Š., Bowker, L., Bytyçi, E., Catalá, A., et al. (2021) · doi:10.17011/jyx/reports/20210518/1 · Cited by 85
‘Just What do You Think You’re Doing, Dave?’ A Checklist for Responsible Data Use in NLP
Rogers, A., Baldwin, T., & Leins, K. (2021) · Findings of the Association for Computational Linguistics: EMNLP 2021 · doi:10.18653/v1/2021.findings-emnlp.414 · Cited by 53
(2024) · doi:10.52202/079017-0970 · Cited by 53
Few-Shot Cross-Lingual Stance Detection with Sentiment-Based Pre-training
Hardalov, M., Arora, A., Nakov, P., & Augenstein, I. (2022) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v36i10.21318 · Cited by 92
“Moreover, when using noisy labels from an external model, there is always a risk of introducing additional bias due to the training data and to discrepancies in the task definition (Waseem et al. 2021; Bender et al. 2021).”
Enabling Sustainable Clouds: The Case for Virtualizing the Energy System
Bashir, N., Guo, T., Hajiesmaili, M., Irwin, D., Shenoy, P. J., Sitaraman, R. K., Souza, A., & Wierman, A. (2021) · ACM Symposium on Cloud Computing (SoCC) · doi:10.1145/3472883.3487009 · Cited by 81
“For example, there has been recent work highlighting deep learning’s carbon emissions [13, 51] and advocating for the design of zero-carbon clouds [17].”
BackgroundRetrieval-Enhanced Machine Learning
Zamani, H., Diaz, F., Dehghani, M., Metzler, D., & Bendersky, M. (2022) · Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval · doi:10.1145/3477495.3531722 · Cited by 52
Probing Pre-Trained Language Models for Cross-Cultural Differences in Values
Arora, A., Kaffee, L.-A., & Augenstein, I. (2023) · Proceedings of the First Workshop on Cross-Cultural Considerations in NLP (C3NLP) · doi:10.18653/v1/2023.c3nlp-1.12 · Cited by 57
“Further, large scale corpora like CommonCrawl over-represent the voices of people with access to the Internet, which in turn over-represents the values of people from those regions (Bender et al., 2021).”
Bob or Bot: Exploring ChatGPT's Answers to University Computer Science Assessment
Richards, M., Waugh, K., Slaymaker, M., Petre, M., Woodthorpe, J., & Gooch, D. (2024) · ACM Transactions on Computing Education · doi:10.1145/3633287 · Cited by 52
“Researchers have described LLMs as (cid:1748) completely mindless(cid:1705) [ 25 ] and (cid:1704)stochastic parrots(cid:1705) [ 6 ].”
BackgroundAI Art is Theft: Labour, Extraction, and Exploitation: Or, On the Dangers of Stochastic Pollocks
Goetze, T. S. (2024) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3630106.3658898 · Cited by 91
Hong, S. J. (2024) · Computers in Human Behavior · doi:10.1016/j.chb.2024.108460 · Cited by 66
Overlap in meaning is a stronger predictor of semantic activation in GPT-3 than in humans
Digutsch, J., & Kosinski, M. (2023) · Scientific Reports · doi:10.1038/s41598-023-32248-6 · Cited by 49
Changing the World by Changing the Data
Rogers, A. (2021) · Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers) · doi:10.18653/v1/2021.acl-long.170 · Cited by 51
“Some dismiss Bender et al. (2021) as “political”, or even “advocacy rather than research” (Lissack, 2021).”
BackgroundInfluentialNeuroLogic A*esque Decoding: Constrained Text Generation with Lookahead Heuristics
Lu, X., Welleck, S., West, P., Jiang, L., Kasai, J., Khashabi, D., Bras, R. J. L., Qin, L., Yu, Y., Zellers, R., et al. (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.57 · Cited by 202
Fair ranking: a critical review, challenges, and future directions
Patro, G. K., Porcaro, L., Mitchell, L., Zhang, Q., Zehlike, M., & Garg, N. (2022) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3531146.3533238 · Cited by 113
Robots Enact Malignant Stereotypes
Hundt, A., Agnew, W. S., Zeng, V., Kacianka, S., & Gombolay, M. (2022) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3531146.3533138 · Cited by 105
“This has only grown in relevance as models and the datasets used to train them have scaled to extremely large, computationally-intensive models [6] that researchers have shown spew racism, sexism, and other forms of harmful bias [6, 14].”
BackgroundMethodInfluentialSpeculative Propositions for Digital Writing Under the New Autonomous Model of Literacy
Robinson, B. (2022) · Postdigital Science and Education · doi:10.1007/s42438-022-00358-5 · Cited by 66
“…2023, which is likely to be followed by At its core, a natural language model is a computational representation of a language, and NLP is the operation by which an algorithm is trained to use a language model to read and write—though, crucially, not to understand (Bender et al. 2021)—a language.”
BackgroundKim, J., Lee, J., Jang, K. M., & Lourentzou, I. (2023) · Telematics and Informatics · doi:10.1016/j.tele.2023.102085 · Cited by 63
The Affective Scaffolding of Grief in the Digital Age: The Case of Deathbots
Fabry, R. E., & Alfano, M. (2024) · Topoi · doi:10.1007/s11245-023-09995-2 · Cited by 72
“This resource is easy to implement and can generate text with virtually no cost or restriction for an individual user – though of course it does require vast compute, energy, and water resources (see Bender et al. 2021).”
Gougherty, A. V., & Clipp, H. L. (2024) · npj Biodiversity · doi:10.1038/s44185-024-00043-9 · Cited by 52
A Critical Analysis of the Largest Source for Generative AI Training Data: Common Crawl
Baack, S. (2024) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3630106.3659033 · Cited by 154
Navigating LLM ethics: advancements, challenges, and future directions
Jiao, J., Afroogh, S., Xu, Y., & Phillips, C. (2025) · AI and Ethics · doi:10.1007/s43681-025-00814-5 · Cited by 219
“Section 2.2 discusses key ethical theories and approaches guiding their responsible development and deployment.”
InfluentialFiD-Light: Efficient and Effective Retrieval-Augmented Text Generation
Hofstätter, S., Chen, J., Raman, K., & Zamani, H. (2023) · Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval · doi:10.1145/3539618.3591687 · Cited by 51
“A major component, besides concerns about safety and social biases as summarized by Bender et al. (2021), is the efficiency of the deployed system.”
BackgroundFrugalGPT: How to Use Large Language Models While Reducing Cost and Improving Performance
Chen, L., Zaharia, M., & Zou, J. (2023) · arXiv · doi:10.48550/arxiv.2305.05176 · Cited by 889
FiD-Light: Efficient and Effective Retrieval-Augmented Text Generation
Hofstätter, S., Chen, J., Raman, K., & Zamani, H. (2022) · arXiv · doi:10.48550/arxiv.2209.14290 · Cited by 51
Sarkar, A. (2023) · Proceedings of the 2nd Annual Meeting of the Symposium on Human-Computer Interaction for Work · doi:10.1145/3596671.3597650 · Cited by 50
Towards education 4.0: The role of large language models as virtual tutors in chemical engineering
Caccavale, F., Gargalo, C. L., Gernaey, K. V., & Krühne, U. (2024) · Education for Chemical Engineers · doi:10.1016/j.ece.2024.07.002 · Cited by 50
Dahlgren Lindström, A., Methnani, L., Krause, L., Ericson, P., de Rituerto de Troya, Í. M., Coelho Mollo, D., & Dobbe, R. (2025) · Ethics and Information Technology · doi:10.1007/s10676-025-09837-2 · Cited by 54
“These models are first trained on large corpora of data such as articles, books, and websites—requiring unprecedented volumes of training data (Bender et al., 2021).”
InfluentialChristoforaki, M., & Beyan, O. (2022) · Applied Sciences · doi:10.3390/app12094130 · Cited by 50
“, in ad targeting [45] and natural language processing (NLP) [46], geo-diversity issues [47], and abusive NLP [48], among others (for a non-academic presentation of impacted sectors see [1], from which the title of this subsection is borrowed).”
BackgroundOn actor-network theory and algorithms: ChatGPT and the new power relationships in the age of AI
Gutiérrez, J. L. M. (2023) · AI and Ethics · doi:10.1007/s43681-023-00314-4 · Cited by 85
Uncovering Energy-Efficient Practices in Deep Learning Training: Preliminary Steps Towards Green AI
Yarally, T., Cruz, L., Feitosa, D., Sallou, J., & van Deursen, A. (2023) · IEEE/ACM International Conference on AI Engineering – Software Engineering for AI (CAIN) · doi:10.1109/cain58948.2023.00012 · Cited by 97
Not My Voice! A Taxonomy of Ethical and Safety Harms of Speech Generators
Hutiri, W. T., Papakyriakopoulos, O., & Xiang, A. (2024) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3630106.3658911 · Cited by 113
Participation in the age of foundation models
Suresh, H., Tseng, E., Young, M., Gray, M., Pierson, E., & Levy, K. (2024) · The 2024 ACM Conference on Fairness Accountability and Transparency · doi:10.1145/3630106.3658992 · Cited by 50
“This gradation aligns with historical perspectives on the co-optation of participation, such as Arnstein’s “Ladder of Citizen Participation,” which describes the degree to which political and economic processes redistribute power to citizens [5].”
BackgroundEvaluating language models for mathematics through interactions
Collins, K. M., Jiang, A. Q., Frieder, S., Wong, L., Zilka, M., Bhatt, U., Lukasiewicz, T., Wu, Y., Tenenbaum, J. B., Hart, W., et al. (2024) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2318124121 · Cited by 49
Improving accuracy of GPT-3/4 results on biomedical data using a retrieval-augmented language model
Soong, D., Sridhar, S., Si, H., Wagner, J.-S., Sá, A. C. C., Yu, C. Y., Karagoz, K., Guan, M., Kumar, S., Hamadeh, H., et al. (2024) · PLOS Digital Health · doi:10.1371/journal.pdig.0000568 · Cited by 49
“A corpus that predominantly features one type of language or cultural context may display bias towards that specific domain or culture [23].”
BackgroundMoëll, B., & Aronsson, F. S. (2025) · Journal of Medical Internet Research · doi:10.2196/75849 · Cited by 71
“However, their capabilities stem from predicting likely sequences of words based on statistical patterns in their training data, not from genuine understanding, reasoning, or consciousness [6].”
A typology of artificial intelligence data work
Muldoon, J., Cant, C., Wu, B., & Graham, M. (2024) · Big Data & Society · doi:10.1177/20539517241232632 · Cited by 50
“These systems are designed with machine learning algorithms that leverage large amounts of data, energy and computational power (Bender et al., 2021; Crawford, 2021).”
Still no lie detector for language models: probing empirical and conceptual roadblocks
Levinstein, B. A., & Herrmann, D. A. (2025) · Philosophical Studies · doi:10.1007/s11098-023-02094-3 · Cited by 50
“3.1 we pushed back against the claims in Shanahan (2022) and Bender et al. (2021) that all an LLM is doing is text prediction, and thus cannot be tracking the truth.”
InfluentialOn the Safety of Conversational Models: Taxonomy, Dataset, and Benchmark
Sun, H., Xu, G., Deng, J., Cheng, J., Zheng, C., Zhou, H., Peng, N., Zhu, X., & Huang, M. (2022) · Findings of the Association for Computational Linguistics: ACL 2022 · doi:10.18653/v1/2022.findings-acl.308 · Cited by 49
Model Positionality and Computational Reflexivity: Promoting Reflexivity in Data Science
Cambo, S. A., & Gergle, D. (2022) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3491102.3501998 · Cited by 49
“The FAccT community spans law, policy, computer science, education, sociology, and philosophy to progress our understanding of how we can develop sociotechnical systems with a core focus on fairness, accountability, and transparency [4, 13, 21, 61].”
BackgroundSengupta, U., & Sengupta, U. (2022) · Frontiers in Sociology · doi:10.3389/fsoc.2022.995603 · Cited by 64
“The areas of concern include increased energy usage by large AI projects (Strubell et al., 2019; Bender et al., 2021), as well as the material impacts of the lifecycle of AI (Crawford, 2021), and political ecnomy of control of AI infrastructure (Whittaker, 2021).”
BackgroundResultInfluentialUnderstanding and Detecting Hallucinations in Neural Machine Translation via Model Introspection
Xu, W., Agrawal, S., Briakou, E., Martindale, M. J., & Carpuat, M. (2023) · Transactions of the Association for Computational Linguistics · doi:10.1162/tacl_a_00563 · Cited by 53
“, 2021), the output is completely detached from the source, which not only reveals fundamental limitations of current models, but also risks misleading users and undermining trust (Bender et al., 2021).”
BackgroundPower and politics in framing bias in Artificial Intelligence policy
Ulnicane, I., & Aden, A. (2023) · Review of Policy Research · doi:10.1111/ropr.12567 · Cited by 72
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.