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 1,201–1,300 of 6,931
Generative-AI-Generated Challenges for Health Data Research
Spector-Bagdady, K. (2023) · The American Journal of Bioethics · doi:10.1080/15265161.2023.2252311 · Cited by 18
“…that by not documenting training data as LLMs are built will lead us to “incurring documentation debt i.e. putting our-selves in a situation where the datasets are both undocumented and too large to document post hoc” which is necessary for accountability and transparency (Bender et al. 2021).”
InfluentialTransforming Agency. On the mode of existence of Large Language Models
Barandiaran, X. E., & Almendros, L. S. (2024) · arXiv · doi:10.48550/arxiv.2407.10735 · Cited by 18
Beyond Detection: Redesigning Authentic Assessment in an AI-Mediated World
Kickbusch, S., Ashford-Rowe, K., Kemp, A., Boreland, J., & Huijser, H. (2025) · Education Sciences · doi:10.3390/educsci15111537 · Cited by 17
University students’ trust in AI: Examining reliance and strategies for critical engagement.
Abuzar, M., Mahmudulhassan, & Muthoifin (2025) · International Journal of Interactive Mobile Technologies (iJIM) · doi:10.3991/ijim.v19i07.52875 · Cited by 20
Laboratory-Scale AI: Open-Weight Models are Competitive with ChatGPT Even in Low-Resource Settings
Wolfe, R., Slaughter, I., Han, B., Wen, B., Yang, Y., Rosenblatt, L., Herman, B., Brown, E., Qu, Z., Weber, N., et al. (2024) · The 2024 ACM Conference on Fairness, Accountability, and Transparency · doi:10.1145/3630106.3658966 · Cited by 24
What's in the Box? A Preliminary Analysis of Undesirable Content in the Common Crawl Corpus
Luccioni, A. S., & Viviano, J. D. (2021) · arXiv · doi:10.48550/arxiv.2105.02732 · Cited by 17
“It will also be important to evaluate whether curation is sufficient: it remains possible that a model could create an undesirable generation from multiple distinct innocuous examples (Bender et al., 2021; Gehman et al., 2020).”
BackgroundOh, D., & Downey, J. (2024) · Information Communication & Society · doi:10.1080/1369118x.2024.2346531 · Cited by 30
“Furthermore, this semi-supervised, lexicon-based approach eliminates the chances of algorithmic biases known in many machine-learning algorithms and hence provides for a level of transparency (Bender et al., 2021).”
Beyond Individual Accountability: (Re-)Asserting Democratic Control of AI
Bogiatzis-Gibbons, D. J. (2024) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3630106.3658541 · Cited by 35
“Through their material embodiment in hardware, LLMs are also part of extractivist supply chains that contribute to climate change and ecosystem degradation [5, 14].”
BackgroundUnderstanding model power in social AI
Brandtzæg, P. B., Skjuve, M. B., & Følstad, A. (2024) · AI & Society · doi:10.1007/s00146-024-02053-4 · Cited by 26
“They can also shape political preferences and public discourse (Rozado 2024), mirror societal biases (Bender et al. 2021), and propagate new norms (De Gregorio 2023), profoundly influencing user interactions and perceptions (Brown et al. 2020).”
Towards Ethical AI Moderation in Multiplayer Games
Sparrow, L. A., Galwey, R., Jovic, D., Hardwick, T., & Butt, M.-A. R. (2024) · Proceedings of the ACM on Human-Computer Interaction · doi:10.1145/3677109 · Cited by 26
“For example, AI systems and processes have been described as “aliens” (for their otherness and inscrutability) [24] , “mulch” [88] (for being collections of “traces of humans”), and “stochastic parrots” (for their meaningless mimicry) [6].”
Human-interpretable clustering of short text using large language models
Miller, J. K., & Alexander, T. J. (2025) · Royal Society Open Science · doi:10.1098/rsos.241692 · Cited by 27
Natural language processing for social science research: A comprehensive review
Hou, Y., & Huang, J. (2025) · Chinese Journal of Sociology · doi:10.1177/2057150x241306780 · Cited by 45
“Numerous studies have demonstrated that the content generated by LLMs inherits and even exacerbates these social biases (Abid et al., 2021; Bender et al., 2021; Gallegos et al., 2024; Kotek et al., 2023; OpenAI et al., 2024).”
Validating the use of large language models for psychological text classification
Bunt, H. L., Goddard, A., Reader, T. W., & Gillespie, A. (2025) · Frontiers in Social Psychology · doi:10.3389/frsps.2025.1460277 · Cited by 31
“Whilst some runs suggested an escalating performance from MinZero to MaxFew, other runs over the same dataset showed the opposite pattern, clearly showcasing the stochastic nature of GPT (Bender et al., 2021).”
Geography for AI sustainability and sustainability for GeoAI
Shi, M., Janowicz, K., Verstegen, J. A., Currier, K., Wiedemann, N., Mai, G., Majić, I., Liu, Z., & Zhu, R. (2025) · Cartography and Geographic Information Science · doi:10.1080/15230406.2025.2479796 · Cited by 29
“There have been increasing calls in the machine learning (ML) community to enhance their energy efficiency (Bender et al., 2021; Schwartz et al., 2020; Strubell et al., 2019).”
Ortloff, A.-M., Martius, F., Meier, M., Sans-Raimbault, T., Geierhaas, L., & Smith, M. (2025) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3713671 · Cited by 30
Angi, A., Sacco, A., & Marchetto, G. (2025) · IEEE Transactions on Network and Service Management · doi:10.1109/tnsm.2025.3570017 · Cited by 38
“Second, our solu-tion works even with a Small Language Model (SLM) for intent interpretation, thus making IBN more sustainable and reducing the energy consumption and the economic expenses of using this type of model that is otherwise resource-hungry [19], [20], [21].”
The Artificial Intelligence Ontology: LLM-Assisted Construction of AI Concept Hierarchies
Joachimiak, M. P., Miller, M. A., Caufield, J. H., Ly, R., Harris, N. L., Tritt, A., Mungall, C., & Bouchard, K. E. (2024) · Applied Ontology · doi:10.1177/15705838241304103 · Cited by 41
“Crucially, we have also included classes related to the ethical and legal impact of AI methods (Bender et al., 2021; Ning et al., 2023; e.g., we provide a top-level Bias class with 61 subclasses).”
Woo, T., Kim, S., Tariq, S., Heo, S., & Yoo, C. (2025) · Korean Journal of Chemical Engineering · doi:10.1007/s11814-025-00524-y · Cited by 22
“Recent studies have explored the use of LLMs for data augmentation and knowledge extraction, enabling automated analysis of technical documentation, experimental reports, and operational logs to uncover meaningful process insights [25].”
Middle school girls using generative AI to engage in mathematical problem-posing
Walkington, C., Pando, M., Lin-Lipsmeyer, L., Beauchamp, T., Sager, M. T., & Milton, S. (2025) · Mathematical Thinking and Learning · doi:10.1080/10986065.2025.2542724 · Cited by 27
Transforming agency: On the mode of existence of large language models
Barandiaran, X. E., & Almendros, L. S. (2025) · Phenomenology and the Cognitive Sciences · doi:10.1007/s11097-025-10094-3 · Cited by 42
“Whereas the metaphor of the “stochastic parrot” was once useful to question the rapidly emerging hype on LLMs (Bender et al., 2021), it might easily turn counterproductive.”
InfluentialLarge Language Models for Psychological Assessment: A Comprehensive Overview
Brickman, J., Gupta, M., & Oltmanns, J. R. (2025) · Advances in Methods and Practices in Psychological Science · doi:10.1177/25152459251343582 · Cited by 50
“LLMs contain biases that are prevalent in society and that researchers and the field at large should be aware of and prepared to continuously address in a transparent manner (Bender et al., 2021).”
InfluentialCountering Racial Bias in Computer Graphics Research
Kim, T., Rushmeier, H., Dorsey, J. O., Nowrouzezahrai, D., Syed, R., Jarosz, W., & Darke, A. M. (2022) · ACM SIGGRAPH Conference (SIGGRAPH) · doi:10.1145/3532836.3536263 · Cited by 24
“Progress is urgently needed, because computer graphics algorithms are increasingly being used to generate training sets for machine learning, so their potential to amplify existing inequalities has never been greater [1].”
BackgroundEmotion Bubbles: Emotional Composition of Online Discourse Before and After the COVID-19 Outbreak
Zhunis, A., Lima, G., Song, H., Han, J., & Cha, M. (2022) · International World Wide Web Conference (WWW) · doi:10.1145/3485447.3512132 · Cited by 23
“It’s Like the Value System in the Loop”: Domain Experts’ Values Expectations for NLP Automation
Showkat, D., & Baumer, E. P. S. (2022) · Designing Interactive Systems Conference · doi:10.1145/3532106.3533483 · Cited by 25
FairCanary: Rapid Continuous Explainable Fairness
Ghosh, A., Shanbhag, A., & Wilson, C. (2022) · Proceedings of the AAAI/ACM Conference on AI Ethics and Society · doi:10.1145/3514094.3534157 · Cited by 35
STRONGHOLD: Fast and Affordable Billion-Scale Deep Learning Model Training
Sun, X., Wang, W., Qiu, S., Yang, R., Huang, S., Xu, J., & Wang, Z. (2022) · International Conference for High Performance Computing, Networking, Storage and Analysis (SC) · doi:10.1109/sc41404.2022.00076 · Cited by 35
Transformers in Natural Language Processing
Yvon, F. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-24349-3_6 · Cited by 25
“As shown on many occasions [Bender et al., 2021], these models in particular tend to amplify the biases present in the data, and may, for example, generate uncontrolled statements of a sexist or racist nature.”
ResultFrom Stochastic Parrots to Intelligent Assistants—The Secrets of Data and Human Interventions
Fayyad, U. M. (2023) · IEEE Intelligent Systems · doi:10.1109/mis.2023.3268723 · Cited by 25
Teaching (with) Artificial Intelligence: The Next Twenty Years
Michels, S. (2023) · Journal of Political Science Education · doi:10.1080/15512169.2023.2266848 · Cited by 17
“Emily M. Bender has deemed this version of AI a “stochastic parrot” (Bender, Gebru, and McMillan-Major 2012).”
Urrutia, F., & Araya, R. (2023) · Journal of Educational Computing Research · doi:10.1177/07356331231191174 · Cited by 32
“However, limitations such as budget constraints, availability, and privacy, among other factors, make it challenging to explore the full potential of GPT models comprehensively (Bender, E. M. et al., 2021).”
BackgroundTechnology hype: Dealing with bold expectations and overpromising
Bareis, J., Roßmann, M., & Bordignon, F. (2023) · TATuP - Zeitschrift für Technikfolgenabschätzung in Theorie und Praxis · doi:10.14512/tatup.32.3.10 · Cited by 28
Epistemic Power in AI Ethics Labor: Legitimizing Located Complaints
Widder, D. G. (2024) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3630106.3658973 · Cited by 45
“Microsoft recently laid off one of its AI ethics teams, defending this decision by saying the central ethics team was broken up and individual workers moved “within the individual product teams that are building the services and the software” [85].”
BackgroundTake a shot! Natural language control of intelligent robotic X-ray systems in surgery
Killeen, B. D., Chaudhary, S., Osgood, G. M., & Unberath, M. (2024) · International Journal of Computer Assisted Radiology and Surgery · doi:10.1007/s11548-024-03120-3 · Cited by 24
“Furthermore, open questions persist regarding the quality and ethical use of data used to train state-of-the-art LLMs, which tend to ingest as much of the internet as possible [31].”
Towards a Benchmark for Scientific Understanding in Humans and Machines
Barman, K. G., Caron, S., Claassen, T., & de Regt, H. W. (2024) · Minds and Machines · doi:10.1007/s11023-024-09657-1 · Cited by 36
“This implies that artifi - cial agents, including Large Language Models (LLMs), should not be dismissed out of hand as being incapable of scientific understanding, simply because they allegedly “guess the next word” or are “stochastic parrots” (Bender et al., 2021 ).”
BackgroundMapping the individual, social and biospheric impacts of Foundation Models
Hernández, A. D., Krishna, S., Perini, A., Katell, M. A., Bennett, S., Borda, A., Hashem, Y., Hadjiloizou, S., Mahomed, S., Jayadeva, S., et al. (2024) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3630106.3658939 · Cited by 48
Large Language Model-based Test Case Generation for GP Agents
Jorgensen, S., Nadizar, G., Pietropolli, G., Manzoni, L., Medvet, E., O’Reilly, U.-M., & Hemberg, E. (2024) · Genetic and Evolutionary Computation Conference (GECCO) · doi:10.1145/3638529.3654056 · Cited by 24
Leite, B. S. (2024) · Journal of Applied Learning & Teaching · doi:10.37074/jalt.2024.7.2.13 · Cited by 33
The Dual Nature of AI in Information Dissemination: Ethical Considerations
Germani, F., Spitale, G., & Biller‐Andorno, N. (2024) · JMIR AI · doi:10.2196/53505 · Cited by 38
Bannister, P., & Carver, M. (2024) · Professional Development in Education · doi:10.1080/19415257.2024.2427873 · Cited by 23
Artificial intelligence chatbots mimic human collective behaviour
He, J. K., Wallis, F. P. S., Gvirtz, A., & Rathje, S. (2024) · British Journal of Psychology · doi:10.1111/bjop.12764 · Cited by 36
“Research has found that LLMs may reproduce biases present in the training data (Crockett & Messeri, 2023; Dillion et al., 2023) and are likely to over-represent Western, high-income cultures (Apicella et al., 2020; Atari et al., 2024; Bender et al., 2021; Facts and Figures 2021, 2021).”
Uncovering Implicit Gender Bias in Narratives through Commonsense Inference
Huang, T., Brahman, F., Shwartz, V., & Chaturvedi, S. (2021) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2021.findings-emnlp.326 · Cited by 52
“To generate stories for our analysis, we use a language model, GPT-2, pre-trained on WebText which has been shown to have potential harms and misuses (Bender et al., 2021).”
Hyperparameter Power Impact in Transformer Language Model Training
Chavannes, L. H. P. D., Kongsbak, M. G. K., Lagermann, T. M. R., & Derczynski, L. (2021) · Workshop on Simple and Efficient Natural Language Processing (SustaiNLP) · doi:10.18653/v1/2021.sustainlp-1.12 · Cited by 20
“Researchers with good ideas may not be able to execute those ideas, given that state-of-the-art results are locked behind large-scale compute (Strubell et al., 2019; Bender et al., 2021).”
BackgroundThe Future of Work is No Work: A Call to Action for Designers in the Abolition of Work
Browne, J., & Green, L. (2022) · CHI Conference Extended Abstracts on Human Factors in Computing Systems (CHI EA) · doi:10.1145/3491101.3516385 · Cited by 23
On the Limitations of Dataset Balancing: The Lost Battle Against Spurious Correlations
Schwartz, R., & Stanovsky, G. (2022) · Findings of the Association for Computational Linguistics: NAACL · doi:10.18653/v1/2022.findings-naacl.168 · Cited by 35
Davis, J. L. (2024) · Australian Journal Of International Affairs · doi:10.1080/10357718.2024.2328293 · Cited by 19
“Without the capacity to self-re fl ect or contextualise, AI o ff ers form without meaning (Bender et al. 2021; Bender and Koller 2020), mimicking, but in no way replicating, the thick and recursive processes of human mentation.”
BackgroundFrom attention economy to cognitive lock-ins
Hansen, M. (2024) · Big Data & Society · doi:10.1177/20539517241275878 · Cited by 22
“This has been called stochastic parroting, referring to the probabilistic stitching together of words ‘ without any reference to meaning ’ (Bender et al., 2021: 617).”
AI, Global Governance, and Digital SovereigntyPreprint
Srivastava, S., & Bullock, J. B. (2024) · SSRN Electronic Journal · doi:10.2139/ssrn.4996387 · Cited by 54
Tajik, A. (2025) · Research Square · doi:10.21203/rs.3.rs-5788599/v1 · Cited by 34
How Artificial Intelligence Differs From Humans in Peer Review
Joachim, M. V., Dodson, T. B., & Laviv, A. (2025) · Journal of Oral and Maxillofacial Surgery · doi:10.1016/j.joms.2025.03.015 · Cited by 22
Abu-Jeyyab, M., Alrosan, S., & Alkhawaldeh, I. (2023) · High Yield Medical Reviews · doi:10.59707/hymrfbya5348 · Cited by 17
Can Large Language Models Safely Address Patient Questions Following Cataract Surgery?
Chowdhury, M., Lim, E., Higham, A., McKinnon, R., Ventoura, N., He, Y., & De Pennington, N. (2023) · Proceedings of the 5th Clinical Natural Language Processing Workshop · doi:10.18653/v1/2023.clinicalnlp-1.17 · Cited by 17
Are Large Language Models Moral Hypocrites? A Study Based on Moral Foundations
Nunes, J. L., Almeida, G. F. C. F., Araujo, M. D., & Barbosa, S. D. J. (2024) · Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society · doi:10.1609/aies.v7i1.31704 · Cited by 17
Data-Efficient Strategies for Expanding Hate Speech Detection into Under-Resourced Languages
Röttger, P., Nozza, D., Bianchi, F., & Hovy, D. (2022) · Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/2022.emnlp-main.383 · Cited by 18
LLMs as Hackers: Autonomous Linux Privilege Escalation Attacks
Happe, A., Kaplan, A., & Cito, J. (2026) · Empirical Software Engineering · doi:10.1007/s10664-025-10758-3 · Cited by 22
“…hand, proponents claim that emergent features arise only with larger model sizes (Kosinski 2023; Bubeck et al. 2023; Wei et al. 2022); on the other hand, proponents claim that smaller models can achieve domain-specific tasks with reduced costs for both training and execution (Bender et al. 2021).”
How Good are LLMs in Generating Personalized Advertisements?
Meguellati, E., Han, L., Bernstein, A., Sadiq, S. W., & Demartini, G. (2024) · International World Wide Web Conference (WWW) · doi:10.1145/3589335.3651520 · Cited by 42
“Their potential for creating personalized content raises both opportunities and ethical concerns, particularly regarding privacy, bias, and mis-information [1].”
Gender Bias in Natural Language Processing and Computer Vision: A Comparative Survey
Bartl, M., Mandal, A., Leavy, S., & Little, S. (2025) · ACM Computing Surveys · doi:10.1145/3700438 · Cited by 48
“…large training sets, such as the Common Crawl Corpus used for training the large language model (LLM) GPT-3 [22] containing text data from the openly accessible Internet, suffers from selection bias due to greater Internet access in more developed countries over-representing their perspective [11].”
InfluentialWester, J., Moghe, B., Winkle, K., & van Berkel, N. (2024) · Proceedings of the ACM on Human-Computer Interaction · doi:10.1145/3687036 · Cited by 24
AI, power and sociolinguistics
Maly, I. (2024) · Journal of Sociolinguistics · doi:10.1111/josl.12681 · Cited by 22
Meaning and understanding in large language models
Havlík, V. (2024) · Synthese · doi:10.1007/s11229-024-04878-4 · Cited by 28
“This elementary example shows how we would be forced to think about the possible understanding of a system within a symbolic approach to AI.”
InfluentialGenerative Artificial Intelligence in Marketing: The Invisible Danger of AI Hallucinations
Yaprak, B. (2024) · Journal of Economy Business and Management · doi:10.7596/jebm.1588897 · Cited by 24
Roumbanis, L. (2025) · Futures · doi:10.1016/j.futures.2025.103538 · Cited by 37
Evolution and Optimization of Language Model Architectures: From Foundations to Future Directions
AlQenaei, Z. M. (2025) · Lecture notes in networks and systems · doi:10.1007/978-981-97-7839-3_16 · Cited by 17
Can AI Outsmart Fake News? Detecting Misinformation with AI Models in Real-Time
Gondwe, G. (2025) · Emerging Media · doi:10.1177/27523543251325902 · Cited by 26
“Models often re fl ect the biases present in their training data, which can lead to unequal detection rates across demographics (Bender et al., 2021).”
From revolution to evolution: What generative AI really means for language learning
Pegrum, M. A. (2025) · Language Teaching · doi:10.1017/s0261444825000151 · Cited by 33
“…the fact that the negative climate and environmental costs of AI development and deployment are disproportionately borne by those in the Global South, alongside marginalised populations in the Global North, who benefit the least from these kinds of technologies (Bender et al., 2021, pp. 612–613).”
Between fact and fairy: tracing the hallucination metaphor in AI discourse
Förster, S., & Skop, Y. (2025) · AI & Society · doi:10.1007/s00146-025-02392-w · Cited by 33
Mohammad, S. M. (2022) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2022.acl-long.573 · Cited by 54
Search results diversification for effective fair ranking in academic search
McDonald, G., Macdonald, C., & Ounis, I. (2021) · Information Retrieval · doi:10.1007/s10791-021-09399-z · Cited by 25
“IR systems and machine learned models can encode and perpetuate any biases that exist in the test collections that they use (Zehlike et al. 2017; Bender et al. 2021).”
BackgroundOpportunities and risks in the use of AI in career development practice
Wilson, M., Robertson, P., Cruickshank, P., & Gkatzia, D. (2024) · Journal of the National Institute for Career Education and Counselling · doi:10.20856/jnicec.4807 · Cited by 27
“The complexity of the system architecture means that even where attempts are made to ‘de-bias’ the dataset, prejudiced outcomes can persist due to interactions with the algorithms, or difficulties in identifying which aspects of the data are creating bias in the dataset (Bender et al., 2021).”
BackgroundDAISY: An Implementation of Five Core Principles for Transparent and Accountable Conversational AI
Wahde, M., & Virgolin, M. (2022) · International Journal of Human-Computer Interaction · doi:10.1080/10447318.2022.2081762 · Cited by 19
“…even though DNNs are remarkable at generating language that may appear human-like, one should not forget that this is the byproduct of their excellent capability to model statistical co-occurrences, and not of any real intelligent understanding of discourse (Bender et al., 2021; Daws, 2020).”
BackgroundLekova, A. K., Tsvetkova, P., Tanev, T. K., Mitrouchev, P., & Kostova, S. (2022) · Journal of Mechatronics and Artificial Intelligence in Engineering · doi:10.21595/jmai.2022.22720 · Cited by 20
A typology for exploring the mitigation of shortcut behaviour
Friedrich, F., Stammer, W., Schramowski, P., & Kersting, K. (2023) · Nature Machine Intelligence · doi:10.1038/s42256-023-00612-w · Cited by 41
Reshi, J. A., & Ali, R. (2023) · International Journal of Interactive Multimedia and Artificial Intelligence · doi:10.9781/ijimai.2023.02.007 · Cited by 23
Surprising Advances in Generative Artificial Intelligence Prompt Amazement—and Worries
Mackenzie, D. N. (2023) · Engineering · doi:10.1016/j.eng.2023.04.004 · Cited by 17
Rethinking Transparency as a Communicative Constellation
Eyert, F., & Lopez, P. (2023) · ACM Conference on Fairness, Accountability, and Transparency (FAccT) · doi:10.1145/3593013.3594010 · Cited by 36
SERENGETI: Massively Multilingual Language Models for Africa
Adebara, I., Elmadany, A., Abdul-Mageed, M., & Inciarte, A. A. (2023) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2023.findings-acl.97 · Cited by 73
Training Data Extraction From Pre-trained Language Models: A Survey
Ishihara, S. (2023) · Workshop on Trustworthy Natural Language Processing (TrustNLP) · doi:10.18653/v1/2023.trustnlp-1.23 · Cited by 84
Language-Family Adapters for Low-Resource Multilingual Neural Machine Translation
Chronopoulou, A., Stojanovski, D., & Fraser, A. (2023) · doi:10.18653/v1/2023.loresmt-1.5 · Cited by 54
“Therefore, this pre-trained model can encode biases that could harm marginalized populations (Bender et al., 2021) and could also be used to translate harmful text.”
BackgroundKarusala, N., Yan, S. D., Rajkumar, N., Victoria, G., & Anderson, R. (2023) · Proceedings of the ACM on Human-Computer Interaction · doi:10.1145/3610210 · Cited by 21
“Their appropriateness for these cases is also not clear given that these models do not process meaning [5, 18].”
BackgroundDemonstrations of the Potential of AI-based Political Issue Polling
Sanders, N. E., Ulinich, A., & Schneier, B. (2023) · Harvard Data Science Review · doi:10.1162/99608f92.1d3cf75d · Cited by 58
AI to Train AI: Using ChatGPT to Improve the Accuracy of a Therapeutic Dialogue System
Gabor-Siatkowska, K., Sowański, M., Rzatkiewicz, R., Stefaniak, I., Kozłowski, M., & Janicki, A. (2023) · Electronics · doi:10.3390/electronics12224694 · Cited by 20
Bai, J. Y. H., Zawacki‐Richter, O., & Müskens, W. (2024) · Turkish Online Journal of Distance Education · doi:10.17718/tojde.1248901 · Cited by 30
Large language models and their possible uses in law
Homoki, P., & Ződi, Z. (2024) · Hungarian Journal of Legal Studies · doi:10.1556/2052.2023.00475 · Cited by 41
Beyond the Hype: Towards a Critical Debate About AI Chatbots in Swedish Higher Education
Pargman, T. C., Sporrong, E., Farazouli, A., & McGrath, C. (2024) · Högre utbildning · doi:10.23865/hu.v14.6243 · Cited by 34
“However, we note that AI tools and systems may have been developed unethically (Bender et al., 2021) in terms of the resources they use (Luccioni et al., 2023), the data they are sourced on (Bender et al., 2021), and their apparent contempt for intellectual property (Grynbaum & Mac, 2023), and this…”
BackgroundInfluentialRyoo, J. J., & Blunt, T. (2024) · ACM Transactions on Computing Education · doi:10.1145/3660645 · Cited by 23
“…and Artificial Intelligence (AI) inventions are specifically harming teenagers, women, people of color, and low-income communities further complicate the challenge of how CS education efforts should move forward in preparing vulnerable populations to study and engage with CS [e.g., 2, 3, 25, 62].”
Biden’s Executive Order on AI and the E.U.’s AI Act: A Comparative Computer-Ethical Analysis
Wörsdörfer, M. (2024) · Philosophy & Technology · doi:10.1007/s13347-024-00765-5 · Cited by 21
“…[intelligence community] and ‘big business’ [data brokers and other ‘surveillance capitalism’ organizations]), and • (Further) undermining the trust in democracy and the rule of law (Ajayi et al., 2023; Bender et al., 2021; Center for AI Safety, 2023; Gebru et al., 2023a, b; Zuboff, 2019).”
Artificial Intelligence Policymaking: An Agenda for Sociological Research
Law, T. L., & McCall, L. (2024) · Socius Sociological Research for a Dynamic World · doi:10.1177/23780231241261596 · Cited by 31
“…contribute to polarization and misinformation in political discourse (e.g., Bail et al. 2018; Rafail, O’Connell, and Sager 2024), and how tech companies’ often uncritical uses of data can further disempower marginalized communities (e.g., Benjamin 2019; Brayne 2020; see also Bender et al. 2021).”
Accelerating AI for science: open data science for science
Lawrence, N. D., & Montgomery, J. (2024) · Royal Society Open Science · doi:10.1098/rsos.231130 · Cited by 26
“Concerns about bias [35], privacy and security [36] also influence how generative AI systems can be adopted responsibly for research [37].”
BackgroundAs’ad, M. (2024) · Cureus · doi:10.7759/cureus.69710 · Cited by 29
“Robust bias detection and mitigation strategies are also indispensable for addressing both explicit and subtle biases that may manifest in language use or problem-solving methodologies [32].”
BackgroundCeron, T., Falk, N., Baric, A., Nikolaev, D., & Padó, S. (2024) · Transactions of the Association for Computational Linguistics · doi:10.1162/tacl_a_00710 · Cited by 80
“…belief systems with a substantial amount of internal consistency, similar to human political individuals – in a word, a political “worldview”, a term which is used for this concept in political science (Ecker et al., 2021) and has also been suggested to apply to LLMs (Bender et al., 2021).”
BackgroundA Reproducible Analysis of Sequential Recommender Systems
Betello, F., Purificato, A., Siciliano, F., Trappolini, G., Bacciu, A., Tonellotto, N., & Silvestri, F. (2024) · IEEE Access · doi:10.1109/access.2024.3522049 · Cited by 33
Language-agnostic Topic Classification for Wikipedia
Johnson, I. L., Gerlach, M., & Sáez-Trumper, D. (2021) · International World Wide Web Conference (WWW) · doi:10.1145/3442442.3452347 · Cited by 26
“They also come out of a recognition of the environmental costs of training and deploying large machine learning models, a growing trend among technology companies that puts an emphasis on complexity and size instead of seeking simpler, more sustainable solutions [3].”
BackgroundA Computational Look at Oral History Archives
Pessanha, F., & Salah, A. A. A. (2021) · Journal on Computing and Cultural Heritage · doi:10.1145/3477605 · Cited by 40
Probing the Robustness of Pre-trained Language Models for Entity Matching
Rastaghi, M. A., Kamalloo, E., & Rafiei, D. (2022) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3511808.3557673 · Cited by 36
Guiding the Release of Safer E2E Conversational AI through Value Sensitive Design
Bergman, A. S., Abercrombie, G., Spruit, S. L., Hovy, D., Dinan, E., Boureau, Y.-L., & Rieser, V. (2022) · Annual Meeting of the Special Interest Group on Discourse and Dialogue · doi:10.18653/v1/2022.sigdial-1.4 · Cited by 33
“…impact of natural language processing and its applications has received increasing attention within the NLP community (e.g. Hovy and Spruit, 2016) with Large Language Models (LLMs) as one of the recent primary targets (e.g. Bender et al., 2021; Bommasani et al., 2021; Weidinger et al., 2021).”
BackgroundEvaluating Machines by their Real-World Language UsePreprint
Zellers, R., Holtzman, A., Clark, E., Qin, L., Farhadi, A., & Choi, Y. (2020) · arXiv (Cornell University) · Cited by 15
Dynaboard: An Evaluation-As-A-Service Platform for Holistic Next-Generation Benchmarking
Ma, Z., Ethayarajh, K., Thrush, T., Jain, S., Wu, L., Jia, R., Potts, C., Williams, A., & Kiela, D. (2021) · neural information processing systems · doi:10.48550/arxiv.2106.06052 · Cited by 89
“, compactness, fairness, efficiency – are clearly important [60, 6, 16, 63, 68, 38, 2, 51, 17].”
BackgroundA Survey of Reasoning with Foundation ModelsPreprint
Sun, J., Zheng, C., Xie, E., Liu, Z., Chu, R., Liu, J., Xu, J., Ding, M., Li, H., Geng, M., et al. (2023) · OSF Preprints (OSF Preprints) · doi:10.31219/osf.io/ac4sp · Cited by 157
Kuai, J., Brantner, C., Karlsson, M., Van Couvering, E., & Romano, S. (2026) · New Media & Society · doi:10.1177/14614448251321162 · Cited by 16
“These performance discrepancies across languages suggest inequality in technology deployment (Bender et al., 2021; Helm et al., 2024) and high-light challenges in establishing accountability mechanisms across different contexts (Wieringa, 2020).”
Herath, S., Bashardoust, A., Bole, Y., & Shrestha, Y. R. (2026) · European Journal of Information Systems · doi:10.1080/0960085x.2026.2616042 · Cited by 16
“This raises concerns that the outputs could lead to undesirable or incorrect practices and, more critically, could diminish human autonomy and creativity in the design process (Bender et al., 2021).”
Language Models As or For Knowledge Bases
Razniewski, S., Yates, A., Kassner, N., & Weikum, G. (2021) · arXiv · doi:10.48550/arxiv.2110.04888 · Cited by 16
“explicit knowledge: Errors made by LMs-as-KBs are not random, but exhibit systematic biases [23, 15] due to frequent values and co-occurrences (including indirect co-occurrences captured latently).”
Background
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.