“Was it “stated” or was it “claimed”?: How linguistic bias affects generative language models
Open accessAbstract
People use language in subtle and nuanced ways to convey their beliefs.For instance, saying claimed instead of said casts doubt on the truthfulness of the underlying proposition, thus representing the author's opinion on the matter.Several works have identified classes of words that induce such framing effects.In this paper, we test whether generative language models are sensitive to these linguistic cues.In particular, we test whether prompts that contain linguistic markers of author bias (e.g., hedges, implicatives, subjective intensifiers, assertives) influence the distribution of the generated text.Although these framing effects are subtle and stylistic, we find qualitative and quantitative evidence that they lead to measurable style and topic differences in the generated text, leading to language that is more polarised (both positively and negatively) and, anecdotally, appears more skewed towards controversial entities and events.
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
Recent work shows that LLMs, when prompted with a defined context, can generate believable non-scripted behaviors that are consistent with this context.
“Recent work [46] shows that LLMs, when prompted with a defined context, can generate believable non-scripted behaviors that are consistent with this context.”
From Not What You've Signed Up For: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection (2023) · cites this paper as [46]
The paper reports that GPT-2 and GPT-3 generate text in response to prompts and that the generations’ high-level topics do not drift from the prompts (section 3.5, page not provided). It does not assess whether the models generate believable, non-scripted behaviors; its findings concern generated text and linguistic framing. (LLM, conf=0.85)
Experimental — accuracy not yet measured
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
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 · see citing works
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 48,293 · see citing works
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 · see citing works
On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜
Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021) · Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency · doi:10.1145/3442188.3445922 · Cited by 6,988 · cited 1× · see citing works
“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).”
BackgroundLoper, E., & Bird, S. (2002) · doi:10.3115/1118108.1118117 · Cited by 3,248 · see citing works
NLTK: The Natural Language ToolkitPreprint
Loper, E., & Bird, S. (2002) · arXiv (Cornell University) · doi:10.48550/arxiv.cs/0205028 · Cited by 1,893 · see citing works
Liu, B., Hu, M., & Cheng, J. (2005) · International World Wide Web Conference (WWW) · doi:10.1145/1060745.1060797 · Cited by 1,612 · see citing works
The Curious Case of Neural Text Degeneration
Holtzman, A., Buys, J., Du, L., Forbes, M., & Choi, Y. (2019) · arXiv · doi:10.48550/arxiv.1904.09751 · Cited by 1,304 · see citing works
Learning extraction patterns for subjective expressions
Riloff, E., & Wiebe, J. (2003) · Proceedings of the 2003 conference on Empirical methods in natural language processing - · doi:10.3115/1119355.1119369 · Cited by 1,059 · see citing works
Metadiscourse : Exploring Interaction in Writing
Hyland, K. (2018) · Bloomsbury Academic eBooks · doi:10.5040/9781350063617 · Cited by 898 · see citing works
Wiebe, J., Wilson, T., Bruce, R., Bell, M., & Martin, M. (2004) · Computational Linguistics · doi:10.1162/0891201041850885 · Cited by 748 · see citing works
RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Models
Gehman, S., Gururangan, S., Sap, M., Choi, Y., & Smith, N. A. (2020) · Findings of the Association for Computational Linguistics: EMNLP 2020 · doi:10.18653/v1/2020.findings-emnlp.301 · Cited by 693 · see citing works
StereoSet: Measuring stereotypical bias in pretrained language models
Nadeem, M., Bethke, A., & Reddy, S. (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.416 · Cited by 643 · see citing works
Universal Adversarial Triggers for Attacking and Analyzing NLP
Wallace, E., Feng, S., Kandpal, N., Gardner, M., & Singh, S. (2019) · Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP) · doi:10.18653/v1/d19-1221 · Cited by 642 · see citing works
Thomas, M., Pang, B., & Lee, L. (2006) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.3115/1610075.1610122 · Cited by 480 · see citing works
(2020) · arXiv (Cornell University) · Cited by 461 · see citing works
Recognizing Stances in Ideological On-Line Debates
Somasundaran, S., & Wiebe, J. (2010) · Cited by 297 · see citing works
Linguistic Models for Analyzing and Detecting Biased Language
Recasens, M. V., Danescu-Niculescu-Mizil, C., & Jurafsky, D. (2013) · Cited by 294 · see citing works
Karttunen, L. (1971) · Language · doi:10.2307/412084 · Cited by 282 · see citing works
HOOPER, J. B. (1975) · BRILL eBooks · doi:10.1163/9789004368828_005 · Cited by 218 · see citing works
Get out the vote: Determining support or opposition from Congressional floor-debate transcriptsPreprint
Thomas, M., Pang, B., & Lee, L. (2006) · arXiv (Cornell University) · doi:10.48550/arxiv.cs/0607062 · Cited by 168 · see citing works
Greene, S., & Resnik, P. (2009) · Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT) · doi:10.3115/1620754.1620827 · Cited by 164 · see citing works
Connotation Lexicon: A Dash of Sentiment Beneath the Surface Meaning
Feng, S., Kang, J. R. W., Кузнецова, П., & Choi, Y. (2013) · Cited by 124 · see citing works
Identifying featured articles in wikipedia
Lipka, N., & Stein, B. (2010) · International World Wide Web Conference (WWW) · doi:10.1145/1772690.1772847 · Cited by 83 · see citing works
Connotation Frames: A Data-Driven Investigation
Rashkin, H., Singh, S., & Choi, Y. (2016) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/p16-1030 · Cited by 79 · see citing works
Contrasting Opposing Views of News Articles on Contentious Issues
Park, S., Lee, K. S., & Song, J. (2011) · Cited by 78 · see citing works
Recognizing Arguing Subjectivity and Argument Tags
Conrad, A., Wiebe, J., & Hwa, R. (2012) · Cited by 33 · see citing works
Pavalanathan, U., Han, X., & Eisenstein, J. (2018) · Proceedings of the ACM on Human-Computer Interaction · doi:10.1145/3274406 · Cited by 22 · see citing works
From Toxicity in Online Comments to Incivility in American News: Proceed with Caution
Hede, A., Agarwal, O., Lu, L., Mutz, D. C., & Nenkova, A. (2021) · Conference of the European Chapter of the Association for Computational Linguistics · doi:10.18653/v1/2021.eacl-main.225 · Cited by 16 · see citing works
Automatic Detection of Point of View Differences in Wikipedia
Al‐Khatib, K., Schütze, H., & Kantner, C. (2012) · University of Groningen research database (University of Groningen / Centre for Information Technology) · Cited by 15 · see citing works
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2018) · arXiv · doi:10.48550/arxiv.1810.04805
Flagged first, then by citations
3 of the 34 references couldn’t be matched to a known paper.
Research works printed but not linked: 4
- Liu, B., Hu, M., and Cheng, J. (2005). Opinion ob- server: analyzing and comparing opinions on the web. In Proceedings of the 14th international con- ference on World Wide Web, pages 342-351.
- Nadeem, M., Bethke, A., and Reddy, S. (2020). Stere- oset: Measuring stereotypical bias in pretrained lan- guage models. arXiv preprint arXiv:2004.09456. · arXiv
- Radford, A., Wu, J., Child, R., Luan, D., Amodei, D., and Sutskever, I. (2019). Language models are un- supervised multitask learners. OpenAI blog, 1(8):9.
- Sennrich, R., Haddow, B., and Birch, A. (2015). Neu- ral machine translation of rare words with subword units. arXiv preprint arXiv:1508.07909. · arXiv
Cited by
Sorted by citations
Greshake, K., Abdelnabi, S., Mishra, S., Endres, C., Holz, T., & Fritz, M. (2023) · Proceedings of the 16th ACM Workshop on Artificial Intelligence and Security · doi:10.1145/3605764.3623985 · Cited by 667 · see citing works
“LLMs might make up facts (“hallucinate”), generate polarized content, or reproduce biases, hate speech, or stereotypes [6, 7, 16, 27, 40, 47, 68].”
BackgroundLanguage Model Behavior: A Comprehensive Survey
Chang, T. A., & Bergen, B. K. (2024) · Computational Linguistics · doi:10.1162/coli_a_00492 · Cited by 105 · see citing works
McIntosh, T. R., Susnjak, T., Liu, T., Watters, P., & Halgamuge, M. N. (2024) · IEEE Transactions on Cognitive and Developmental Systems · doi:10.1109/tcds.2024.3377445 · Cited by 62 · see citing works
Navigating the Grey Area: How Expressions of Uncertainty and Overconfidence Affect Language Models
Zhou, K., Jurafsky, D., & Hashimoto, T. (2023) · Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/2023.emnlp-main.335 · Cited by 38 · see citing works
Prompt-to-SQL Injections in LLM-Integrated Web Applications: Risks and Defenses
Pedro, R., Coimbra, M. E., Castro, D., Carreira, P., & Santos, N. (2025) · International Conference on Software Engineering · doi:10.1109/icse55347.2025.00007 · Cited by 56 · see citing works
Entity-Based Evaluation of Political Bias in Automatic Summarization
Zhou, K., & Tan, C. (2023) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2023.findings-emnlp.696 · Cited by 5 · see citing works
Bias Neutralization in Non-Parallel Texts: A Cyclic Approach with Auxiliary Guidance
Madanagopal, K., & Caverlee, J. (2023) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2023.emnlp-main.882 · Cited by 2 · see citing works
Zero-Shot Classification by Logical Reasoning on Natural Language Explanations
Han, C., Pei, H., Du, X., & Ji, H. (2023) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2023.findings-acl.571 · Cited by 2 · see citing works
Kořínek, M., & Štekerová, K. (2026) · Journal of Cases on Information Technology · doi:10.4018/jcit.398628 · Cited by 1 · see citing works
A Comprehensive Review of Security and Privacy Issues in Large Language ModelsPreprint
Inamdar, A. S., & Eswaran, S. (2024) · SSRN Electronic Journal · doi:10.2139/ssrn.4825655 · Cited by 1 · see citing works
Conscientization, Dialogue, and Praxis
Warr, M., Yalnaty, M. E., Chatterjee, S., Hour, N., Oster, N. J., & Peterson, S. (2026) · Advances in computational intelligence and robotics book series · doi:10.4018/979-8-3373-7902-9.ch001
- Authors
- Patel, R., & Pavlick, E.
- Affiliations
- Brown University (US)
- Venue
- Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing
- Published
- 2021
- Pages
- 10080-10095
- OpenAlex
- W3212275118
- Cited by
- 11
- References
- 34
- Citations by year
- 2026: 2 · 2025: 1 · 2024: 2 · 2023: 6
- FWCI
- 0.92
- Type
- conference paper
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- cc-by
- Locations
- Conference on Empirical Methods in Natural Language Processing (EMNLP) (publishedVersion, cc-by)
- Funders
- National Science Foundation (1956221)
- Microsoft Research
- Intelligence Advanced Research Projects Activity
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Sentiment Analysis and Opinion Mining
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Physical Sciences › Computer Science › Artificial Intelligence › Natural Language Processing Techniques
- Keywords
- language bias, framing effects, author bias, hedges, implicatives, assertive speech acts, generative language models, prompt effects, text style, topic distribution, political polarization
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
