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 1,501–1,600 of 6,931
The need for an empirical research program regarding human–AI relational norms
Reinecke, M. G., Kappes, A., Mann, S. P., Savulescu, J., & Earp, B. D. (2025) · AI and Ethics · doi:10.1007/s43681-024-00631-2 · Cited by 22
Towards AI-Assisted Mapmaking: Assessing the Capabilities of GPT-4o in Cartographic Design
Memduhoğlu, A. (2025) · ISPRS International Journal of Geo-Information · doi:10.3390/ijgi14010035 · Cited by 16
Bonnet, S., & Teuteberg, F. (2025) · International Journal of Knowledge Management · doi:10.4018/ijkm.368223 · Cited by 18
Designing LLM Chains by Adapting Techniques from Crowdsourcing Workflows
Grunde-McLaughlin, M., Lam, M. S., Krishna, R., Weld, D. S., & Heer, J. (2025) · ACM Transactions on Computer-Human Interaction · doi:10.1145/3716134 · Cited by 45
“For example, LLMs struggle to output pluralistic responses [140], are limited to the outputs of people with access to the Internet [16], and cannot represent identity groups [148].”
Maclure, J., & Morin-Martel, A. (2025) · Digital Society · doi:10.1007/s44206-025-00174-x · Cited by 19
The voice of artificial intelligence: Philosophical and educational reflections
Jackson, L., Sidorkin, A. M., Jandrić, P., Costello, E., Heybach, J. A., Greenhalgh‐Spencer, H., Hytten, K., Gourlay, L. J., Buchanan, R., & Tesař, M. (2025) · Educational Philosophy and Theory · doi:10.1080/00131857.2025.2481588 · Cited by 14
AI in Neurology: Everything, Everywhere, All at Once Part 1: Principles and Practice
Rizzo, M., & Dawson, J. D. (2025) · Annals of Neurology · doi:10.1002/ana.27225 · Cited by 20
“43 2020s Popularization of the term “ stochastic parrots ” by Bender et al 44 highlighting limitations and ethical considerations of large language models; rise of AI-powered conversational agents like ChatGPT and Claude.”
Aplicaciones de inteligencia artificial generativa en la transformación digital empresarial
Galarza-Sánchez, P. C., Boné-Andrade, M. F., & Bravo, V. J. P. (2023) · Revista Científica Ciencia y Método · doi:10.55813/gaea/rcym/v1/n1/8 · Cited by 25
Integrating Computer Prediction Methods in Social Science: A Comment on Hofman et al. (2021)
Breznau, N. (2022) · Social Science Computer Review · doi:10.1177/08944393211049776 · Cited by 12
“Still more concerns relate to the environmental impact of computer energy consumption in larger and larger predictive models (Bender et al., 2021) and evidence that humans often can predict outcomes just as well as machine learning algorithms in sociological and psychological studies (Christodoulou et al.”
BackgroundIncorporating Stock Market Signals for Twitter Stance Detection
Conforti, C., Berndt, J., Pilehvar, M. T., Giannitsarou, C., Toxvaerd, F., & Collier, N. (2022) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2022.acl-long.281 · Cited by 21
“Moreover, the BerTweet model we are using as main text encoder might encode biases due to the data it was trained on (Bender et al., 2021).”
MethodWhat company do words keep? Revisiting the distributional semantics of J.R. Firth & Zellig Harris
Brunila, M., & LaViolette, J. (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.327 · Cited by 32
Tackling Documentation Debt: A Survey on Algorithmic Fairness Datasets
Fabris, A., Messina, S., Silvello, G., & Susto, G. A. (2022) · ACM Conference on Equity and Access in Algorithms, Mechanisms, and Optimization (EAAMO) · doi:10.1145/3551624.3555286 · Cited by 24
“In recent work, Bender et al. [9] propose the notion of documentation debt , in relation to training sets that are undocumented and too large to document retrospectively.”
BackgroundNovel Applications for VAE-based Anomaly Detection Systems
Bergamin, L., Carraro, T., Polato, M., & Aiolli, F. (2022) · IEEE International Joint Conference on Neural Network · doi:10.1109/ijcnn55064.2022.9892879 · Cited by 13
So, C., Khvan, A., & Choi, W. (2023) · Computer Animation and Virtual Worlds · doi:10.1002/cav.2149 · Cited by 20
COBRA Frames: Contextual Reasoning about Effects and Harms of Offensive Statements
Zhou, X., Zhu, H., Yerukola, A., Davidson, T., Hwang, J. D., Swayamdipta, S., & Sap, M. (2023) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2023.findings-acl.392 · Cited by 60
McQuillan, D., Jarke, J., & Pargman, T. C. (2023) · Postdigital Science and Education · doi:10.1007/s42438-023-00433-5 · Cited by 18
On the Unexpected Abilities of Large Language Models
Nolfi, S. (2024) · Adaptive Behavior · doi:10.1177/10597123241256754 · Cited by 24
“This is realized by embedding the LLM within a turn taking system that interleaves model-generated text with user-generated text, and by prompting the LLM with text that induces the model to initiate a conversation with the user.”
Interrogating Human-centered Data Science: Taking Stock of Opportunities and Limitations
Tanweer, A., Aragon, C., Muller, M. J., Guha, S., Passi, S., Neff, G., & Kogan, M. (2022) · CHI Conference Extended Abstracts on Human Factors in Computing Systems (CHI EA) · doi:10.1145/3491101.3503740 · Cited by 15
Emerging trends: Unfair, biased, addictive, dangerous, deadly, and insanely profitable
Church, K., Schoene, A. M., Ortega, J. E., Chandrasekar, R., & Kordoni, V. (2022) · Natural Language Engineering · doi:10.1017/s1351324922000481 · Cited by 23
Fairness and Discrimination in Information Access SystemsPreprint
Ekstrand, M. D., Das, A., Burke, R., & Díaz, F. (2021) · arXiv (Cornell University) · Cited by 13
Societal Biases in Language Generation: Progress and ChallengesPreprint
Sheng, E., Chang, K.-W., Natarajan, P., & Peng, N. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2105.04054 · Cited by 12
“Furthermore, by filtering out all words deemed “bad”, Bender et al. (2021) warns that we remove the discourse of marginalized populations.”
BackgroundLarge Language Models as Simulated Economic Agents: What Can We Learn from Homo Silicus?Preprint
Horton, J. J., Filippas, A., & Manning, B. S. (2023) · SSRN Electronic Journal · doi:10.2139/ssrn.4413859 · Cited by 41
Barker, F. G., & Rutka, J. T. (2023) · Journal of neurosurgery · doi:10.3171/2023.4.jns23482 · Cited by 12
Tell me what you don’t know: large language models and the pathologies of intelligence analysis
Logan, S. (2024) · Australian Journal Of International Affairs · doi:10.1080/10357718.2024.2331733 · Cited by 16
“LLMs su ff er from issues of algorithmic bias shared with other forms of AI (Bender et al .”
BackgroundPotentially Perilous Preference Parrots: Why Digital Twins Do Not Respect Patient Autonomy
Starke, G., & Jox, R. J. (2024) · The American Journal of Bioethics · doi:10.1080/15265161.2024.2353810 · Cited by 13
“Representing purely associative, stochastic models, LLMs have famously been compared with parrots, processing vast language-based data on a probabilistic basis without discernible reference to the texts’ meaning (Bender et al. 2021).”
BackgroundPolitical Ideology and Polarization: A Multi-dimensional Approach
Sinno, B., Oviedo, B., Atwell, K., Alikhani, M., & Li, J. 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.17 · Cited by 29
Supermind Ideator: Exploring generative AI to support creative problem-solving
Rick, S. R., Giacomelli, G., Wen, H., Laubacher, R. J., Taubenslag, N., Heyman, J. L., Knicker, M. S., Jeddi, Y., Maier, H., Dwyer, S., et al. (2023) · arXiv · doi:10.48550/arxiv.2311.01937 · Cited by 12
“A complex topic concerning LLMs today is their safety and accuracy with generated content [3, 44].”
BackgroundHybrid LLM: Cost-Efficient and Quality-Aware Query Routing
Ding, D., Mallick, A., Wang, C., Sim, R., Mukherjee, S., Ruhle, V., Lakshmanan, L. V. S., & Awadallah, A. H. (2024) · arXiv · doi:10.48550/arxiv.2404.14618 · Cited by 12
Sharma, S., Klemettilä, P. A., & Junko, T. (2025) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3713204 · Cited by 16
“Similarly, previous research has shown the prevalence of belief in AI as a mysterious or "fuzzy" concept that nevertheless works as a "silver bullet" for solving complex social problems, often encouraged by over-enthusiastic industry promises and sensationalist media coverage [7, 48, 51, 53, 55].”
AI and the Future of Language Teaching: Motivating Sustained Integrated Professional Development
Hubbard, P. L., & Schulze, M. (2025) · International Journal of Computer-Assisted Language Learning and Teaching · doi:10.4018/ijcallt.378304 · Cited by 24
Benz, P., Pradier, C., Kozlowski, D., Shokida, N. S., & Larivière, V. (2025) · Scientometrics · doi:10.1007/s11192-025-05339-6 · Cited by 22
“Within this context, we face the risk of not knowing the exact origin of the data used for training (Bender et al., 2021) introduce bias (Nemani et al., 2023).”
Kaiser, C., Kaiser, J., Manewitsch, V., Rau, L., & Schallner, R. (2025) · ACM Conference on User Modeling, Adaptation and Personalization (UMAP) · doi:10.1145/3708319.3733685 · Cited by 44
“One concern in the generation of synthetic data with LLMs is the potential for sociodemographic bias, meaning systematically lower accuracy for certain population subgroups, e.g., due to limitations or imbalances in the LLM’s training data [6, 24].”
Fairness Beyond the Algorithmic Frame: Actionable Recommendations for an Intersectional Approach
Vethman, S., Smit, Q. T. S., van Liebergen, N. M., & Veenman, C. J. (2025) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3715275.3732210 · Cited by 18
Shvetsova, O. A., Katalshov, D., & Lee, S.-K. (2025) · Applied Sciences · doi:10.3390/app15137298 · Cited by 26
Maruyama, H., Toyama, Y., Takanami, K., Takase, K., & Kamei, T. (2025) · JMIR Medical Education · doi:10.2196/69313 · Cited by 12
Tang, B., Sun, K., Curran, N. T., Schaub, F., & Shin, K. G. (2025) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3770640 · Cited by 49
“Not only are they error-prone, generating inaccurate, biased, or misleading information [13, 30, 52, 76], LLMs also struggle with hallucinating content [39].”
Kumar, V., Kotler, P., & Kumar, A. (2025) · Journal of the Academy of Marketing Science · doi:10.1007/s11747-025-01120-6 · Cited by 17
Uncoupling inequality: Reflections on the ethics of benchmarks for digital media
Washington, A. L., Rhue, L. A., Nakamura, L., & Stevens, R. C. (2022) · Hawaii International Conference on System Sciences (HICSS) · doi:10.24251/hicss.2022.352 · Cited by 22
Learning to Predict User-Defined Types
Jesse, K., Dévanbu, P., & Sawant, A. A. (2022) · IEEE Transactions on Software Engineering · doi:10.1109/tse.2022.3178945 · Cited by 19
“Pre-training on large corpora is compute-intensive, which is often performed at large, resource-rich organizations that can afford the cost of training [25], [26].”
BackgroundExposure and emergence in usage-based grammar: computational experiments in 35 languages
Dunn, J. (2022) · Cognitive Linguistics · doi:10.1515/cog-2021-0106 · Cited by 16
Improving Gender-Related Fairness in Sentence Encoders: A Semantics-Based Approach
Dolci, T., Azzalini, F., & Tanelli, M. (2023) · Data Science and Engineering · doi:10.1007/s41019-023-00211-0 · Cited by 23
Which Stereotypes Are Moderated and Under-Moderated in Search Engine Autocompletion?
Leidinger, A., & Rogers, R. (2023) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3593013.3594062 · Cited by 21
How Accurate are GPT-3’s Hypotheses About Social Science Phenomena?
Rosenbusch, H., Stevenson, C. E., & van der Maas, H. L. J. (2023) · Digital Society · doi:10.1007/s44206-023-00054-2 · Cited by 23
“However, especially for ethically-loaded hypotheses, an LLM based on word associations in large internet scrapes should not be given much credibility (cf., Bender et al., 2021; Corbett-Davies & Goel, 2018).”
Voice in the Machine: Ethical Considerations for Language-Capable Robots
Williams, T., Matuszek, C., Jokinen, K., Korpan, R., Pustejovsky, J., & Scassellati, B. (2023) · Communications of the ACM · doi:10.1145/3604632 · Cited by 11
“…as smart speakers (such as encoding and perpetuation of hegemonically dominant white heteropatriarchal stereotypes, norms, and biases 6 and climate risks), 1 but they also present fundamentally new and accentuated risks that stem from the confluence of their communicative capability and embodiment.”
BackgroundMichaelov, J. A., & Bergen, B. K. (2023) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2023.findings-acl.891 · Cited by 18
Automated Text Generation and Summarization for Academic Writing
Benites, F., Benites, A. D., & Anson, C. M. (2023) · Springer eBooks · doi:10.1007/978-3-031-36033-6_18 · Cited by 21
Yokoyama, A., Ferro, M., & Schulze, B. R. (2023) · AI Communications · doi:10.3233/aic-230063 · Cited by 18
Jermakowicz, E. K. (2023) · Journal of global awareness · doi:10.24073/jga/4/02/03 · Cited by 21
Nothing Is Harder to Resist Than the Temptation of AI
Park, A., Kietzmann, J., Killoran, J., Cui, Y. G., van Esch, P., & Dabirian, A. (2023) · IT Professional · doi:10.1109/mitp.2023.3340529 · Cited by 13
“Some of these near-term risks have been identi fi ed more generally by other researchers (e.g., Bender et al., 6 ) and we draw upon Kapoor and Narayanan ’ s related work 7 on this topic.”
BackgroundLarge language models and academic writing: Five tiers of engagement
Bekker, M. (2024) · South African Journal of Science · doi:10.17159/sajs.2024/17147 · Cited by 29
iScore: Visual Analytics for Interpreting How Language Models Automatically Score Summaries
Coscia, A. J., Holmes, L., Morris, W., Choi, J. S., Crossley, S. A., & Endert, A. (2024) · ACM International Conference on Intelligent User Interfaces (IUI) · doi:10.1145/3640543.3645142 · Cited by 19
“Transparency is critical for building trust in using LLMs [30], especially for learning engineers who are increasingly using LLMs as “black boxes” for downstream tasks, as well as during human-in-the-loop evaluation of LLM performance where quantitative benchmarks often fall short [3].”
BackgroundBrunet-Gouet, É., Vidal, N., & Roux, P. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-55245-8_7 · Cited by 13
Manifestations of xenophobia in AI systems
Tomašev, N., Maynard, J. L., & Gabriel, I. (2024) · AI & Society · doi:10.1007/s00146-024-01893-4 · Cited by 25
“The overall impact of AI on the job market spans beyond its immediate utilisation in employment decisions, and should be seen in relation to how technology is reshaping the future of work—including job quality—across industries and regions.”
Butcher, B., Zilka, M., Hron, J., Cook, D., & Weller, A. (2024) · ACM Journal on Responsible Computing · doi:10.1145/3652591 · Cited by 18
“, 2022), although not without limitations and biases (Bender et al., 2021).”
BackgroundComputation, data and AI in Anthropocene history
Wickberg, A., & Gärdebo, J. (2023) · History and Technology · doi:10.1080/07341512.2024.2330724 · Cited by 15
Gould, S. J. J. (2024) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642206 · Cited by 15
“Data-driven science is often the norm, in contrast to Tay-lorism’s demand for planned, controlled experimentation.”
The extimate core of understanding: absolute metaphors, psychosis and large language models
Heimann, M., & Hübener, A.-F. (2024) · AI & Society · doi:10.1007/s00146-024-01971-7 · Cited by 15
“The probabilistic nature of LLMs is repeated in most publications on them, and the derisive “stochastic parrot” introduced by Bender et al. (2021) informs this reading.”
Constructing Capabilities: The Politics of Testing Infrastructures for Generative AI
Grill, G. (2024) · Conference on Fairness, Accountability and Transparency · doi:10.1145/3630106.3659009 · Cited by 21
“This, in practice, often means adhering to a hegemonic white individualist male standard recognized as “normal” by many customers [11, 102].”
BackgroundNLP for Maternal Healthcare: Perspectives and Guiding Principles in the Age of LLMs
Antoniak, M., Naik, A., Alvarado, C. S., Wang, L. L., & Chen, I. Y. (2024) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3630106.3658982 · Cited by 51
“Within the fairness, accountability, and trans-parency (FAccT) community, researchers have studied many general risks and challenges arising with LLMs [9, 103] and also high-lighted risks specific to healthcare settings.”
BackgroundAkal Badi ya Bias: An Exploratory Study of Gender Bias in Hindi Language Technology
Hada, R., Husain, S., Gumma, V., Diddee, H., Yadavalli, A., Seth, A., Kulkarni, N., Gadiraju, U., Vashistha, A., Seshadri, V., et al. (2024) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3630106.3659017 · Cited by 33
A Survey of Text-Matching Techniques
Jiang, P., & Cai, X. (2024) · Information · doi:10.3390/info15060332 · Cited by 26
Information and library professionals' roles and responsibilities in an AI ‐augmented world
Marchionini, G. (2024) · Journal of the Association for Information Science and Technology · doi:10.1002/asi.24930 · Cited by 12
“…also a host of papers, blogs, and conference workshops that critically discuss the technical limitations and practical implications of GenAI (e.g., Bender et al., 2021; Kapoor & Narayanan's, n.d. substack book AI Snake Oil, https://www.aisnakeoil.com/; or the Marcus's, n.d. on AI substack,…”
Things to Do in the Translation Class When Technologies Change: The Case of Generative AI
Ayvazyan, N., Hao, Y., & Pym, A. (2024) · New frontiers in translation studies · doi:10.1007/978-981-97-2958-6_11 · Cited by 17
Sadeghi, M., Pöttgen, D., Ebel, P., & Vogelsang, A. (2024) · ACM Conference on User Modeling, Adaptation and Personalization (UMAP) · doi:10.1145/3627043.3659573 · Cited by 26
“Given the current progress in AI, we expect even more persuasive explanations in the future that may even increase the explanation bias [6, 10, 24, 26, 27, 33, 34].”
BackgroundHirsbrunner, S. D. (2024) · Wiley Interdisciplinary Reviews Climate Change · doi:10.1002/wcc.902 · Cited by 17
“At the same time, research (Bender et al., 2021) has shown that LLMs overrepresent hegemonic languages, cultures, and communities (e.g., English, US, and social media users), while others are marginalized (notably from the Global South).”
ChatGPT in Higher Education: Perceptions of Computer Science-Related Students
Hammer, S., Ottinger, S., Zönnchen, B., Hohendanner, M., Hobelsberger, M., & Thurner, V. (2024) · IEEE Global Engineering Education Conference (EDUCON) · doi:10.1109/educon60312.2024.10578742 · Cited by 15
“By confusing NLP with NLU the “seemingly coherent language [generated by LLMs] heightens the potential for automation bias, deliberate misuse, and amplification of a hegemonic worldview” [16].”
BackgroundPrompt Engineering for Applied Linguistics: Elements, Examples, Techniques, and Strategies
Lin, Z. (2024) · English Language Teaching · doi:10.5539/elt.v17n9p14 · Cited by 22
“For instance, Bender et al. (2021) revealed that LLMs like GPT-3, when prompted with gender-neutral phrases, would often produce gender-skewed outputs.”
BackgroundUsing Generative AI and ChatGPT for improving the production of distance learning materials
Mikroyannidis, A., Sharma, N., Ekuban, A., & Domingue, J. (2024) · IEEE International Conference on Advanced Learning Technologies (ICALT) · doi:10.1109/icalt61570.2024.00061 · Cited by 14
Greer, K. M., Khan, S., Sangmo, D., Greene, A. R., & Rutkowski, D. (2024) · Sex Education · doi:10.1080/14681811.2024.2401802 · Cited by 14
“However, if users are unfamiliar with bias in AI responses or AI hallucination, users may not be discouraged from using ChatGPT and may assume information is accurate (Bender et al. 2021).”
Generative AI Search Engines as Arbiters of Public Knowledge: An Audit of Bias and Authority
Li, A., & Sinnamon, L. (2024) · Proceedings of the Association for Information Science and Technology · doi:10.1002/pra2.1021 · Cited by 15
“Such techniques pose the danger that bad actors may take advantage of these outputs as LLMs are not subject to truth or ethics (Bender et al., 2021).”
InfluentialAuditLLM: A Tool for Auditing Large Language Models Using Multiprobe Approach
Amirizaniani, M., Martin, E., Roosta, T. G., Chadha, A. R., & Shah, C. S. (2024) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3627673.3679222 · Cited by 33
Transformative landscape of ChatGPT in higher education: unveiling the academic discourse
Shrivastava, A. (2024) · Global Knowledge, Memory and Communication · doi:10.1108/gkmc-02-2024-0068 · Cited by 15
Why ChatGPT doesn’t think: an argument from rationality
Stoljar, D., & Zhang, Z. V. (2024) · Inquiry · doi:10.1080/0020174x.2024.2427061 · Cited by 12
“…and Cappelen 2023; Lederman and Mahowald 2024; Yildirim and Paul 2024) argue as if it is, others argue that ChatGPT is no more than a ‘ stochastic parrot ’ , a thing that mimics thinking within limits rather than doing the real thing (Bender et al. 2021; see also Bender and Koller 2020; Go ff 2023).”
Contextualizing predictive minds
Butz, M. V., Mittenbühler, M., Schwöbel, S., Achimova, A., Gumbsch, C., Otte, S., & Kiebel, S. J. (2024) · Neuroscience & Biobehavioral Reviews · doi:10.1016/j.neubiorev.2024.105948 · Cited by 22
“For example, as reviewed above, LLMs mostly fail in slightly elaborate reasoning tasks (Bender et al., 2021; Binz and Schulz, 2023).”
InfluentialSupermind 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) · Collective Intelligence · doi:10.1177/26339137241305117 · Cited by 14
Unlocking Sustainability Compliance: Characterizing the EU Taxonomy for Business Process Management
Klessascheck, F., Fahrenkrog-Petersen, S. A., Mendling, J., & Pufahl, L. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-78338-8_18 · Cited by 12
“Moreover, technical limitations of LLMs, such as “confabulations”, “hallucinations” and inherent biases (see, e.g., [4,26,42]), apply to our study as well.”
BackgroundHuang, H., Yang, J., & Ren, C. (2025) · Journal of the Knowledge Economy · doi:10.1007/s13132-025-02640-w · Cited by 15
“Strubell et al. (2019) and Bender et al. (2021) demonstrate that energy consumption and carbon emissions from training AI models adversely affect green growth.”
Exploring Multiverses: Generative AI and Neuroaesthetic Perspectives
Forte, M. (2025) · Heritage · doi:10.3390/heritage8030102 · Cited by 14
“One primary challenge is the risk of bias, mistakes and misrepresentation—AI models are only as good as the data they are trained on [23], and historical and archaeological records (and their interpretations) might reflect the use of incorrect data, potentially marginalizing alternative narratives.”
Toward Language Justice: Exploring Multilingual Captioning for Accessibility
Desai, A., Alharbi, R., Hsueh, S., Ladner, R. E., & Mankoff, J. C. (2025) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3713622 · Cited by 23
“Some technologies actively erase non-normative language practices: e.g., Bender et al’s work highlights how language models flatten variation and contribute to reifying linguistic hierarchies [12].”
Laflamme, A. S., & Bruneault, F. (2025) · Journal of Scholarly Publishing · doi:10.3138/jsp-2024-1125 · Cited by 48
“The production of biased content by generative AI models poses a significant issue for academia, as these systems often reproduce and amplify the prejudices and stereotypes embedded in their training data (Bender et al., 2021 ; Buolamwini & Gebru, 2018 ; Nicoletti & Bass, 2023 ).”
Addressing Ethical Considerations and Responsible AI Practices
Luthra, A., Dixit, S., Garg, S., Singh, A., & Anchal, M. (2025) · Emerald Group Publishing Limited eBooks · doi:10.1108/978-1-83549-852-120251007 · Cited by 13
Balancing Large Language Model Alignment and Algorithmic Fidelity in Social Science Research
Lyman, A., Hepner, B., Argyle, L. P., Busby, E. C., Gubler, J. R., & Wingate, D. (2025) · Sociological Methods & Research · doi:10.1177/00491241251342008 · Cited by 33
“…human counterparts, models are most helpful and do the least harm when they are free as much as possible from algorithmic bias and the misinformed, prejudiced, or toxic information/speech that results from it (Bender et al. 2021; Caliskan et al. 2017; Kleinberg et al. 2018; Obermeyer et al. 2019).”
What Does a Software Engineer Look Like? Exploring Societal Stereotypes in LLMs
Bano, M., Gunatilake, H., & Hoda, R. (2025) · IEEE/ACM International Conference on Software Engineering: Software Engineering in Society (ICSE-SEIS) · doi:10.1109/icse-seis66351.2025.00023 · Cited by 29
No NLP Task Should be an Island: Multi-disciplinarity for Diversity in News Recommender Systems
Reuver, M., Fokkens, A. S., & Verberne, S. (2021) · Leiden Repository (Leiden University) · Cited by 25
“Was it “stated” or was it “claimed”?: How linguistic bias affects generative language models
Patel, R., & Pavlick, E. (2021) · Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/2021.emnlp-main.790 · Cited by 11
“Societal biases in language models Several recent works have looked at bias in language models and the societal effects they may have (Bender et al., 2021; Nadeem et al., 2020).”
BackgroundHate Speech Detection using OpenAI and GPT-3
Gupta, S. K. (2022) · International Journal of Emerging Technology and Advanced Engineering · doi:10.46338/ijetae0522_15 · Cited by 15
Robertson, S., & Díaz, M. (2022) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3531146.3534638 · Cited by 31
“they reflect anymeaning or intent [4].”
BackgroundLahnala, A., WELCH, C. A., Neuendorf, B., & Flek, L. (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.363 · Cited by 25
The Ethics of Computational Social Science
Leslie, D. (2023) · Springer eBooks · doi:10.1007/978-3-031-16624-2_4 · Cited by 20
TESCREAL hallucinations: Psychedelic and AI hype as inequality engines
Devenot, N. (2023) · Journal of Psychedelic Studies · doi:10.1556/2054.2023.00292 · Cited by 22
Intuition as a “trained thing”: sensing, thinking, and speculating in computational cultures
Pedwell, C. (2023) · Subjectivity · doi:10.1057/s41286-023-00170-x · Cited by 20
An Exploratory Literature Study on Sharing and Energy Use of Language Models for Source Code
Hort, M., Grishina, A., & Moonen, L. (2023) · Proceedings of the ... ACM-IEEE International Symposium on Empirical Software Engineering and Measurement/International Symposium on Empirical Software Engineering and Measurement · doi:10.1109/esem56168.2023.10304803 · Cited by 18
Prompt2Model: Generating Deployable Models from Natural Language Instructions
Viswanathan, V., Zhao, C., Bertsch, A., Wu, T., & Neubig, G. (2023) · Conference on Empirical Methods in Natural Language Processing: System Demonstrations (EMNLP) · doi:10.18653/v1/2023.emnlp-demo.38 · Cited by 59
A Blueprint for an AI Bill of Rights for Education
Conrad, K. (2024) · Critical AI · doi:10.1215/2834703x-11205245 · Cited by 14
Fine-Grained Fairness Analysis of Abusive Language Detection Systems with CheckList
Manerba, M. M., & Tonelli, S. (2021) · Workshop on Online Abuse and Harms (WOAH) · doi:10.18653/v1/2021.woah-1.9 · Cited by 25
“Although BERT and similar language models may already encode biases (Bender et al., 2021), fine-tuning on different datasets may indeed lead to a change in classification behaviour and therefore in its implicit biases.”
BackgroundBenchmarking Resource Usage for Efficient Distributed Deep Learning
Frey, N. C., Li, B., McDonald, J., Zhao, D., Jones, M. P., Bestor, D., Tiwari, D., Gadepally, V. N., & Samsi, S. (2022) · IEEE High Performance Extreme Computing Conference · doi:10.1109/hpec55821.2022.9926375 · Cited by 23
“As deep learning (DL) workflows become more prevalent in the sciences, neural architecture search and hyperparameter sweeps consume an increasingly enormous amount of compute and power resources [2], [4], [26], [27], [35], [40], [44] at high-performance computing (HPC) centers [1], [37], [42] and cloud providers.”
BackgroundDExperts: Decoding-Time Controlled Text Generation with Experts and Anti-ExpertsPreprint
Liu, A., Sap, M., Lu, X., Swayamdipta, S., Bhagavatula, C., Smith, N. A., & Choi, Y. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2105.03023 · Cited by 10
“For a broader discussion of such risks, and of the risks of large pretrained LMs in general, please see Bender et al. (2021).”
BackgroundWho's in and who's out? A case study of multimodal CLIP-filtering in DataComp
Hong, R., Agnew, W., Kohno, T., & Morgenstern, J. (2024) · Proceedings of the 4th ACM Conference on Equity and Access in Algorithms, Mechanisms, and Optimization · doi:10.1145/3689904.3694702 · Cited by 11
Collaborative Storytelling with Human Actors and AI Narrators
Branch, B., Mirowski, P., & Mathewson, K. W. (2021) · arXiv · doi:10.48550/arxiv.2109.14728 · Cited by 11
“Live curation, mitigating of model bias There are however trade-offs between the predictive power of large language models, and their embedded biases or their misalignment with desired societal values, which have been discussed in (Bender et al. 2021; Kenton et al. 2021).”
BackgroundDeanthropomorphising NLP: Can a language model be conscious?
Shardlow, M., & Przybyła, P. (2024) · PLoS ONE · doi:10.1371/journal.pone.0307521 · Cited by 16
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.