---
title: "Whose Opinions Do Language Models Reflect? — Santurkar 2023"
description: "Language models increasingly answer subjective questions that have no correct answer, so any reply, including a refusal, takes a stance."
authors: ["Shibani Santurkar", "Esin Durmus", "Faisal Ladhak", "Cinoo Lee", "Percy Liang", "Tatsunori Hashimoto"]
year: 2023
venue: "arXiv:2303.17548"
source: "https://arxiv.org/abs/2303.17548"
doi: "10.48550/arxiv.2303.17548"
arxiv: "2303.17548"
licence: "arXiv — non-exclusive license to distribute"
licence_url: "http://arxiv.org/licenses/nonexclusive-distrib/1.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/santurkar-2023-whose-opinions-language-models-reflect/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/santurkar-2023-whose-opinions-language-models-reflect/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2023](https://md.research.elamur.ai/free/en/papers/2023/page/8/) / Whose Opinions Do Language Models Reflect?

# Whose Opinions Do Language Models Reflect?

**Authors:** Shibani Santurkar, [Esin Durmus](https://md.research.elamur.ai/free/en/authors/esin-durmus/), [Faisal Ladhak](https://md.research.elamur.ai/free/en/authors/faisal-ladhak/), Cinoo Lee, [Percy Liang](https://md.research.elamur.ai/free/en/authors/percy-liang/), [Tatsunori Hashimoto](https://md.research.elamur.ai/free/en/authors/tatsunori-hashimoto/) · **Year:** 2023 · **Venue:** arXiv:2303.17548 · **Licence:** [arXiv — non-exclusive license to distribute](http://arxiv.org/licenses/nonexclusive-distrib/1.0/)

**Links:** [Publisher's page](https://arxiv.org/abs/2303.17548) · [PDF](https://arxiv.org/pdf/2303.17548) · [DOI 10.48550/arxiv.2303.17548](https://doi.org/10.48550/arxiv.2303.17548) · [arXiv 2303.17548](https://arxiv.org/abs/2303.17548) · [OpenAlex](https://openalex.org/W4361866126)

**Citations:** 89 citations (openalex-snapshot · 2026-10-14) · [301](#cited-by) in our library, of the 140884 papers whose bibliographies we read

## Summary

Language models increasingly answer subjective questions that have no correct answer, so any reply, including a refusal, takes a stance. The authors ask which human groups' views these stances match. To measure this, they turn professionally designed public opinion polls into an evaluation resource. The resulting dataset, OpinionQA, contains 1498 multiple-choice questions from Pew Research's American Trends Panel. Each question comes with weighted answer distributions for the overall US population and for 60 demographic groups.

For each question, the model's next-token log probabilities over the answer options are normalized into an opinion distribution and compared with the human distributions. The comparison covers three axes:
- **representativeness**: default agreement with groups
- **steerability**: whether persona prompts shift answers toward a group
- **consistency**: whether group alignment holds across topics

Across nine models, the gap between model and population views is about as large as the gap between Democrats and Republicans on climate change. Human-feedback tuning widens this gap. Persona prompting helps only slightly. The authors present OpinionQA as a diagnostic probe rather than a target to optimize.

## Key points

- **Data and models:** OpinionQA draws on 15 ATP survey waves, covering topics such as privacy, politics, health, guns and religion. Nine models from OpenAI and AI21 Labs (350M–178B parameters) were evaluated.
- **Shift after human-feedback tuning:** Tuned models align more with liberal, highly educated and higher-income respondents. text-davinci-003 often collapses onto a single dominant view, for example giving Joe Biden over 99% approval.
- **Poorly represented groups:** All models represent people aged 65+, Mormons and widowed individuals poorly.
- **Steering is limited:** Steering prompts (QA, BIO, PORTRAY formats) improve group alignment only modestly and fix none of the representativeness gaps.
- **Inconsistent leanings and scope:** Alignment is not stable across topics. Generally liberal text-davinci-002/003 give conservative-leaning answers on religion. Conclusions apply only to the US, and the multiple-choice results may not carry over to open-ended generation.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

Fields: [Other](https://md.research.elamur.ai/free/en/shelves/other/)

## Cited by

Cited by

301 papers of our library cite it (of the 140884 whose bibliographies we read); 13 of them have a page here, 288 not yet.

- [Bridging Cultural Nuances in Dialogue Agents through Cultural Value Surveys](https://md.research.elamur.ai/free/en/papers/cao-2024-bridging-cultural-nuances-dialogue-agents/)
- [From Persona to Personalization: A Survey on Role-Playing Language Agents](https://md.research.elamur.ai/free/en/papers/chen-2024-persona-personalization-survey-role-playing/)
- [A Rose by Any Other Name: LLM-Generated Explanations Are Good Proxies for Human Explanations to Collect Label Distributions on NLI](https://md.research.elamur.ai/free/en/papers/chen-2025-rose-any-other-name-llm/)
- [Marked Personas: Using Natural Language Prompts to Measure Stereotypes in Language Models](https://md.research.elamur.ai/free/en/papers/cheng-2023-marked-personas-natural-language-prompts/)
- [Towards Measuring the Representation of Subjective Global Opinions in Language Models](https://md.research.elamur.ai/free/en/papers/durmus-2023-towards-measuring-representation-subjective-global/)
- [Bridging the Gap: A Survey on Integrating (Human) Feedback for Natural Language Generation](https://md.research.elamur.ai/free/en/papers/fernandes-2023-bridging-gap-survey-integrating-human/)
- [Not what you've signed up for: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection](https://md.research.elamur.ai/free/en/papers/greshake-2023-ve-signed-up-compromising-real/)
- [Having Beer after Prayer? Measuring Cultural Bias in Large Language Models](https://md.research.elamur.ai/free/en/papers/naous-2023-having-beer-after-prayer-measuring/)
- [Political Compass or Spinning Arrow? Towards More Meaningful Evaluations for Values and Opinions in Large Language Models](https://md.research.elamur.ai/free/en/papers/rottger-2024-political-compass-or-spinning-arrow/)
- [German General Social Survey Personas: A Survey-Derived Persona Prompt Collection for Population-Aligned LLM Studies](https://md.research.elamur.ai/free/en/papers/rupprecht-2026-german-general-social-survey-personas/)
- [Children's English Reading Story Generation via Supervised Fine-Tuning of Compact LLMs with Controllable Difficulty and Safety](https://md.research.elamur.ai/free/en/papers/shen-2026-children-s-english-reading-story/)
- [A Roadmap to Pluralistic Alignment](https://md.research.elamur.ai/free/en/papers/sorensen-2024-roadmap-pluralistic-alignment/)
- [Can Large Language Models Transform Computational Social Science?](https://md.research.elamur.ai/free/en/papers/ziems-2023-large-language-models-transform-computational/)

## Cite

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@misc{santurkar2023whose,
  title = {{Whose Opinions Do Language Models Reflect?}},
  author = {Santurkar, Shibani and Durmus, Esin and Ladhak, Faisal and Lee, Cinoo and Liang, Percy and Hashimoto, Tatsunori},
  year = {2023},
  doi = {10.48550/arxiv.2303.17548},
  eprint = {2303.17548},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2303.17548}
}
```

- [← previous When Your Language Model Cannot Even Do Determiners Right: Probing for Anti-Presuppositions and the Maximize Presupposition! Principle](https://md.research.elamur.ai/free/en/papers/sieker-2023-when-your-language-model-cannot/)
- [next → Why Can GPT Learn In-Context? Language Models Secretly Perform Gradient Descent as Meta-Optimizers](https://md.research.elamur.ai/free/en/papers/dai-2023-why-can-gpt-learn-context/)
- [↑ back to papers of 2023](https://md.research.elamur.ai/free/en/papers/2023/page/8/)

## References

41 references, 13 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Aher, G., Arriaga, R., and Kalai, A. Using large language models to simulate multiple humans. arXiv preprint arXiv:2208.10264, 2022. — [arXiv:2208.10264](https://arxiv.org/abs/2208.10264) [find](https://scholar.google.com/scholar?q=Aher%2C+G%2E%2C+Arriaga%2C+R%2E%2C+and+Kalai%2C+A%2E+Using+large+language+models+to+simulate+multiple+humans%2E+arXiv+preprint+arXiv%3A2208%2E10264%2C+2022%2E+%E2%80%94+%5BarXiv%3A2208%2E10264%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2208%2E10264%29)
2. AI21Labs. Jurassic-1 Instruct [beta]. [https://docs.ai21.com/docs/jurassic-1-instruct-beta](https://docs.ai21.com/docs/jurassic-1-instruct-beta), 2022. [↗](https://docs.ai21.com/docs/jurassic-1-instruct-beta)
3. Alm, C. Subjective natural language problems: Motivations, applications, characterizations, and implications. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, 2011. [find](https://scholar.google.com/scholar?q=Alm%2C+C%2E+Subjective+natural+language+problems%3A+Motivations%2C+applications%2C+characterizations%2C+and+implications%2E+In+Proceedings+of+the+49th+Annual+Meeting+of+the+Association+for+Computational+Linguistics)
4. Argyle, L. P., Busby, E. C., Fulda, N., Gubler, J., Rytting, C., and Wingate, D. Out of one, many: Using language models to simulate human samples. arXiv preprint arXiv:2209.06899, 2022. [find](https://scholar.google.com/scholar?q=Argyle%2C+L%2E+P%2E%2C+Busby%2C+E%2E+C%2E%2C+Fulda%2C+N%2E%2C+Gubler%2C+J%2E%2C+Rytting%2C+C%2E%2C+and+Wingate%2C+D%2E+Out+of+one%2C+many%3A+Using+language+models+to+simulate+human+samples%2E+arXiv+preprint+arXiv%3A2209%2E06899%2C+2022%2E) [our card](https://md.research.elamur.ai/free/en/papers/argyle-2022-out-one-many-using-language/)
5. Askell, A., Bai, Y., Chen, A., Drain, D., Ganguli, D., Henighan, T., Jones, A., Joseph, N., en, B. M., DasSarma, N., et al. A general language assistant as a laboratory for alignment. arXiv preprint arXiv:2112.00861, 2021. [find](https://scholar.google.com/scholar?q=Askell%2C+A%2E%2C+Bai%2C+Y%2E%2C+Chen%2C+A%2E%2C+Drain%2C+D%2E%2C+Ganguli%2C+D%2E%2C+Henighan%2C+T%2E%2C+Jones%2C+A%2E%2C+Joseph%2C+N%2E%2C+en%2C+B%2E+M%2E%2C+DasSarma%2C+N%2E%2C+et+al%2E+A+general+language+assistant+as+a+laboratory+for+alignment%2E+arXiv+preprint+a)
6. Bai, Y., Kadavath, S., Kundu, S., Askell, A., Kernion, J., Jones, A., Chen, A., Goldie, A., Mirhoseini, A., McKinnon, C., et al. Constitutional ai: Harmlessness from ai feedback. arXiv preprint arXiv:2212.08073, 2022. [find](https://scholar.google.com/scholar?q=Bai%2C+Y%2E%2C+Kadavath%2C+S%2E%2C+Kundu%2C+S%2E%2C+Askell%2C+A%2E%2C+Kernion%2C+J%2E%2C+Jones%2C+A%2E%2C+Chen%2C+A%2E%2C+Goldie%2C+A%2E%2C+Mirhoseini%2C+A%2E%2C+McKinnon%2C+C%2E%2C+et+al%2E+Constitutional+ai%3A+Harmlessness+from+ai+feedback%2E+arXiv+preprint+arXiv%3A) [our card](https://md.research.elamur.ai/free/en/papers/bai-2022-constitutional-ai-harmlessness-ai-feedback/)
7. Berinsky, A. Measuring public opinion with surveys. Annual review of political science, 2017. [find](https://scholar.google.com/scholar?q=Berinsky%2C+A%2E+Measuring+public+opinion+with+surveys%2E+Annual+review+of+political+science%2C+2017%2E)
8. Binz, M. and Schulz, E. arXiv:2206.14576, 2022. — [arXiv:2206.14576](https://arxiv.org/abs/2206.14576) [find](https://scholar.google.com/scholar?q=Binz%2C+M%2E+and+Schulz%2C+E%2E+arXiv%3A2206%2E14576%2C+2022%2E+%E2%80%94+%5BarXiv%3A2206%2E14576%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2206%2E14576%29)
9. Using cognitive psychology to understand gpt-3. Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., Agarwal, S., Herbert-Voss, A., Krueger, G., Henighan, T., Child, R., Ramesh, A., Ziegler, D. M., Wu, J., Winter, C., Hesse, C.… [find](https://scholar.google.com/scholar?q=Using+cognitive+psychology+to+understand+gpt%2D3%2E+Brown%2C+T%2E+B%2E%2C+Mann%2C+B%2E%2C+Ryder%2C+N%2E%2C+Subbiah%2C+M%2E%2C+Kaplan%2C+J%2E%2C+Dhariwal%2C+P%2E%2C+Neelakantan%2C+A%2E%2C+Shyam%2C+P%2E%2C+Sastry%2C+G%2E%2C+Askell%2C+A%2E%2C+Agarwal%2C+S%2E%2C+Herbert%2DVoss%2C)
10. Davani, D., Dı́az, M., and Prabhakaran, V. Dealing with disagreements: Looking beyond the majority vote in subjective annotations. Transactions of the Association for Computational Linguistics, 2022. [find](https://scholar.google.com/scholar?q=Davani%2C+D%2E%2C+D%C4%B1%CC%81az%2C+M%2E%2C+and+Prabhakaran%2C+V%2E+Dealing+with+disagreements%3A+Looking+beyond+the+majority+vote+in+subjective+annotations%2E+Transactions+of+the+Association+for+Computational+Linguistics%2C+2022%2E) [our card](https://md.research.elamur.ai/free/en/papers/davani-2022-dealing-disagreements-looking-beyond-majority/)
11. De-Arteaga, M., Romanov, A., Wallach, H., Chayes, J., Borgs, C., Chouldechova, A., Geyik, S., Kenthapadi, K., and Kalai, A. T. Bias in bios: A case study of semantic representation bias in a high-stakes setting. In Conference on Fairness, Accountability, and Transparency, FAT* ’19, pp. 120–128, New… [find](https://scholar.google.com/scholar?q=De%2DArteaga%2C+M%2E%2C+Romanov%2C+A%2E%2C+Wallach%2C+H%2E%2C+Chayes%2C+J%2E%2C+Borgs%2C+C%2E%2C+Chouldechova%2C+A%2E%2C+Geyik%2C+S%2E%2C+Kenthapadi%2C+K%2E%2C+and+Kalai%2C+A%2E+T%2E+Bias+in+bios%3A+A+case+study+of+semantic+representation+bias+in+a+high%2Dstak)
12. Dhamala, J., Sun, T., Kumar, V., Krishna, S., Pruksachatkun, Y., Chang, K., and Gupta, R. BOLD: Dataset and metrics for measuring biases in open-ended language generation. In ACM Conference on Fairness, Accountability, and Transparency, FAccT ’21, pp. 862–872, New York, NY, USA, 2021. Association fo… [find](https://scholar.google.com/scholar?q=Dhamala%2C+J%2E%2C+Sun%2C+T%2E%2C+Kumar%2C+V%2E%2C+Krishna%2C+S%2E%2C+Pruksachatkun%2C+Y%2E%2C+Chang%2C+K%2E%2C+and+Gupta%2C+R%2E+BOLD%3A+Dataset+and+metrics+for+measuring+biases+in+open%2Dended+language+generation%2E+In+ACM+Conference+on+Fairnes)
13. Ganguli, D., Lovitt, L., Kernion, J., Askell, A., Bai, Y., Kadavath, S., Mann, B., Perez, E., Schiefer, N., Ndousse, K., et al. Red teaming language models to reduce harms: Methods, scaling behaviors, and lessons learned. arXiv preprint arXiv:2209.07858, 2022. [find](https://scholar.google.com/scholar?q=Ganguli%2C+D%2E%2C+Lovitt%2C+L%2E%2C+Kernion%2C+J%2E%2C+Askell%2C+A%2E%2C+Bai%2C+Y%2E%2C+Kadavath%2C+S%2E%2C+Mann%2C+B%2E%2C+Perez%2C+E%2E%2C+Schiefer%2C+N%2E%2C+Ndousse%2C+K%2E%2C+et+al%2E+Red+teaming+language+models+to+reduce+harms%3A+Methods%2C+scaling+behaviors%2C) [our card](https://md.research.elamur.ai/free/en/papers/ganguli-2022-red-teaming-language-models-reduce/)
14. Gao, L., Tow, J., Biderman, S., Black, S., DiPofi, A., Foster, C., Golding, L., Hsu, J., McDonell, K., Muennighoff, N., Phang, J., Reynolds, L., Tang, E., Thite, A., Wang, B., Wang, K., and Zou, A. A framework for few-shot language model evaluation, September 2021. URL https://doi.org/10.5281/zenod… — [doi:10.5281/zenodo.5371628](https://doi.org/10.5281/zenodo.5371628) [↗](https://doi.org/10.5281/zenod…)
15. Glaese, A., McAleese, N., Trebacz, M., Aslanides, J., Firoiu, V., Ewalds, T., Rauh, M., Weidinger, L., Chadwick, M., Thacker, P., et al. Improving alignment of dialogue agents via targeted human judgements. arXiv preprint arXiv:2209.14375, 2022. [find](https://scholar.google.com/scholar?q=Glaese%2C+A%2E%2C+McAleese%2C+N%2E%2C+Trebacz%2C+M%2E%2C+Aslanides%2C+J%2E%2C+Firoiu%2C+V%2E%2C+Ewalds%2C+T%2E%2C+Rauh%2C+M%2E%2C+Weidinger%2C+L%2E%2C+Chadwick%2C+M%2E%2C+Thacker%2C+P%2E%2C+et+al%2E+Improving+alignment+of+dialogue+agents+via+targeted+human+judge) [our card](https://md.research.elamur.ai/free/en/papers/glaese-2022-improving-alignment-dialogue-agents-targeted/)
16. Gordon, M., Zhou, K., Patel, K., Hashimoto, T. B., and Bernstein, M. The disagreement deconvolution: Bringing machine learning performance metrics in line with reality. In Conference on Human Factors in Computing Systems (CHI), 2021. arXiv preprint Gordon, M., Lam, M., Park, J., Patel, K., Hancock,… [find](https://scholar.google.com/scholar?q=Gordon%2C+M%2E%2C+Zhou%2C+K%2E%2C+Patel%2C+K%2E%2C+Hashimoto%2C+T%2E+B%2E%2C+and+Bernstein%2C+M%2E+The+disagreement+deconvolution%3A+Bringing+machine+learning+performance+metrics+in+line+with+reality%2E+In+Conference+on+Human+Factors+)
17. Goyal, N., Kivlichan, I., Rosen, R., and Vasserman, L. Is your toxicity my toxicity? exploring the impact of rater identity on toxicity annotation. Proceedings of the ACM on Human-Computer Interaction, 2022. [find](https://scholar.google.com/scholar?q=Goyal%2C+N%2E%2C+Kivlichan%2C+I%2E%2C+Rosen%2C+R%2E%2C+and+Vasserman%2C+L%2E+Is+your+toxicity+my+toxicity%3F+exploring+the+impact+of+rater+identity+on+toxicity+annotation%2E+Proceedings+of+the+ACM+on+Human%2DComputer+Interaction)
18. Guo, C., Pleiss, G., Sun, Y., and Weinberger, K. Q. On calibration of modern neural networks. In International Conference on Machine Learning (ICML), pp. 1321–1330, 2017. [find](https://scholar.google.com/scholar?q=Guo%2C+C%2E%2C+Pleiss%2C+G%2E%2C+Sun%2C+Y%2E%2C+and+Weinberger%2C+K%2E+Q%2E+On+calibration+of+modern+neural+networks%2E+In+International+Conference+on+Machine+Learning+%28ICML%29%2C+pp%2E+1321%E2%80%931330%2C+2017%2E)
19. Hartmann, J., Schwenzow, J., and Witte, M. The political ideology of conversational ai: Converging evidence on chatgpt’s pro-environmental, left-libertarian orientation. arXiv preprint arXiv:2301.01768, 2023. [find](https://scholar.google.com/scholar?q=Hartmann%2C+J%2E%2C+Schwenzow%2C+J%2E%2C+and+Witte%2C+M%2E+The+political+ideology+of+conversational+ai%3A+Converging+evidence+on+chatgpt%E2%80%99s+pro%2Denvironmental%2C+left%2Dlibertarian+orientation%2E+arXiv+preprint+arXiv%3A2301%2E0176)
20. Hendrycks, D., Burns, C., Basart, S., Zou, A., Mazeika, M., Song, D., and Steinhardt, J. Measuring massive multitask language understanding. arXiv preprint arXiv:2009.03300, 2020. [find](https://scholar.google.com/scholar?q=Hendrycks%2C+D%2E%2C+Burns%2C+C%2E%2C+Basart%2C+S%2E%2C+Zou%2C+A%2E%2C+Mazeika%2C+M%2E%2C+Song%2C+D%2E%2C+and+Steinhardt%2C+J%2E+Measuring+massive+multitask+language+understanding%2E+arXiv+preprint+arXiv%3A2009%2E03300%2C+2020%2E) [our card](https://md.research.elamur.ai/free/en/papers/hendrycks-2020-measuring-massive-multitask-language-understanding/)
21. Henrich, J., Heine, S., and Norenzayan, A. The weirdest people in the world? Behavioral and brain sciences, 2010. [find](https://scholar.google.com/scholar?q=Henrich%2C+J%2E%2C+Heine%2C+S%2E%2C+and+Norenzayan%2C+A%2E+The+weirdest+people+in+the+world%3F+Behavioral+and+brain+sciences%2C+2010%2E)
22. Jiang, F., Xu, F., Araki, J., and Neubig, G. How can we know what language models know? Transactions of the Association for Computational Linguistics, 2020. [find](https://scholar.google.com/scholar?q=Jiang%2C+F%2E%2C+Xu%2C+F%2E%2C+Araki%2C+J%2E%2C+and+Neubig%2C+G%2E+How+can+we+know+what+language+models+know%3F+Transactions+of+the+Association+for+Computational+Linguistics%2C+2020%2E) [our card](https://md.research.elamur.ai/free/en/papers/jiang-2020-how-can-we-know-what/)
23. Jiang, H., Beeferman, D., Roy, B., and Roy, D. Communitylm: Probing partisan worldviews from language models. arXiv preprint arXiv:2209.07065, 2022. [find](https://scholar.google.com/scholar?q=Jiang%2C+H%2E%2C+Beeferman%2C+D%2E%2C+Roy%2C+B%2E%2C+and+Roy%2C+D%2E+Communitylm%3A+Probing+partisan+worldviews+from+language+models%2E+arXiv+preprint+arXiv%3A2209%2E07065%2C+2022%2E)
24. Kambhatla, G., Stewart, I., and Mihalcea, R. Surfacing racial stereotypes through identity portrayal. In 2022 ACM Conference on Fairness, Accountability, and Transparency, pp. 1604–1615, 2022. [find](https://scholar.google.com/scholar?q=Kambhatla%2C+G%2E%2C+Stewart%2C+I%2E%2C+and+Mihalcea%2C+R%2E+Surfacing+racial+stereotypes+through+identity+portrayal%2E+In+2022+ACM+Conference+on+Fairness%2C+Accountability%2C+and+Transparency%2C+pp%2E+1604%E2%80%931615%2C+2022%2E)
25. Karra, S., Nguyen, S., and Tulabandhula, T. Ai personification: Estimating the personality of language models. arXiv preprint arXiv:2204.12000, 2022. — [arXiv:2204.12000](https://arxiv.org/abs/2204.12000) [find](https://scholar.google.com/scholar?q=Karra%2C+S%2E%2C+Nguyen%2C+S%2E%2C+and+Tulabandhula%2C+T%2E+Ai+personification%3A+Estimating+the+personality+of+language+models%2E+arXiv+preprint+arXiv%3A2204%2E12000%2C+2022%2E+%E2%80%94+%5BarXiv%3A2204%2E12000%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2204%2E12)
26. Liang, P., Bommasani, R., Lee, T., Tsipras, D., Soylu, D., Yasunaga, M., Zhang, Y., Narayanan, D., Wu, Y., Kumar, A., et al. Holistic evaluation of language models. arXiv preprint arXiv:2211.09110, 2022. [find](https://scholar.google.com/scholar?q=Liang%2C+P%2E%2C+Bommasani%2C+R%2E%2C+Lee%2C+T%2E%2C+Tsipras%2C+D%2E%2C+Soylu%2C+D%2E%2C+Yasunaga%2C+M%2E%2C+Zhang%2C+Y%2E%2C+Narayanan%2C+D%2E%2C+Wu%2C+Y%2E%2C+Kumar%2C+A%2E%2C+et+al%2E+Holistic+evaluation+of+language+models%2E+arXiv+preprint+arXiv%3A2211%2E09110%2C+20)
27. Lieber, O., Sharir, O., Lenz, B., and Shoham, Y. Jurassic-1: Technical details and evaluation. White Paper. AI21 Labs, 2021. [find](https://scholar.google.com/scholar?q=Lieber%2C+O%2E%2C+Sharir%2C+O%2E%2C+Lenz%2C+B%2E%2C+and+Shoham%2C+Y%2E+Jurassic%2D1%3A+Technical+details+and+evaluation%2E+White+Paper%2E+AI21+Labs%2C+2021%2E)
28. Lourie, N., Bras, R. L., and Choi, Y. Scruples: A corpus of community ethical judgments on 32,000 real-life anecdotes. In Proceedings of the AAAI Conference on Artificial Intelligence, 2021. [find](https://scholar.google.com/scholar?q=Lourie%2C+N%2E%2C+Bras%2C+R%2E+L%2E%2C+and+Choi%2C+Y%2E+Scruples%3A+A+corpus+of+community+ethical+judgments+on+32%2C000+real%2Dlife+anecdotes%2E+In+Proceedings+of+the+AAAI+Conference+on+Artificial+Intelligence%2C+2021%2E)
29. Nadeem, M., Bethke, A., and Reddy, S. Stereoset: Measuring stereotypical bias in pretrained language models. arXiv preprint arXiv:2004.09456, 2020. [find](https://scholar.google.com/scholar?q=Nadeem%2C+M%2E%2C+Bethke%2C+A%2E%2C+and+Reddy%2C+S%2E+Stereoset%3A+Measuring+stereotypical+bias+in+pretrained+language+models%2E+arXiv+preprint+arXiv%3A2004%2E09456%2C+2020%2E) [our card](https://md.research.elamur.ai/free/en/papers/nadeem-2021-stereoset-measuring-stereotypical-bias-pretrained/)
30. Ouyang, L., Wu, J., Jiang, X., Almeida, D., Wainwright, C., Mishkin, P., Zhang, C., Agarwal, S., Slama, K., Ray, A., et al. Training language models to follow instructions with human feedback. arXiv preprint arXiv:2203.02155, 2022. [find](https://scholar.google.com/scholar?q=Ouyang%2C+L%2E%2C+Wu%2C+J%2E%2C+Jiang%2C+X%2E%2C+Almeida%2C+D%2E%2C+Wainwright%2C+C%2E%2C+Mishkin%2C+P%2E%2C+Zhang%2C+C%2E%2C+Agarwal%2C+S%2E%2C+Slama%2C+K%2E%2C+Ray%2C+A%2E%2C+et+al%2E+Training+language+models+to+follow+instructions+with+human+feedback%2E+arXiv+p)
31. Park, J. S., Popowski, L., Cai, C., Morris, M. R., Liang, P., and Bernstein, M. S. Social simulacra: Creating populated prototypes for social computing systems. In ACM Symposium on User Interface Software and Technology, 2022. [find](https://scholar.google.com/scholar?q=Park%2C+J%2E+S%2E%2C+Popowski%2C+L%2E%2C+Cai%2C+C%2E%2C+Morris%2C+M%2E+R%2E%2C+Liang%2C+P%2E%2C+and+Bernstein%2C+M%2E+S%2E+Social+simulacra%3A+Creating+populated+prototypes+for+social+computing+systems%2E+In+ACM+Symposium+on+User+Interface+Soft)
32. Pavlick, E. and Kwiatkowski, T. Inherent disagreements in human textual inferences. Transactions of the Association for Computational Linguistics, 2019. [find](https://scholar.google.com/scholar?q=Pavlick%2C+E%2E+and+Kwiatkowski%2C+T%2E+Inherent+disagreements+in+human+textual+inferences%2E+Transactions+of+the+Association+for+Computational+Linguistics%2C+2019%2E) [our card](https://md.research.elamur.ai/free/en/papers/pavlick-2019-inherent-disagreements-human-textual-inferences/)
33. Perez, E., Huang, S., Song, F., Cai, T., Ring, R., Aslanides, J., Glaese, A., McAleese, N., and Irving, G. Red teaming language models with language models. arXiv preprint arXiv:2202.03286, 2022a. [find](https://scholar.google.com/scholar?q=Perez%2C+E%2E%2C+Huang%2C+S%2E%2C+Song%2C+F%2E%2C+Cai%2C+T%2E%2C+Ring%2C+R%2E%2C+Aslanides%2C+J%2E%2C+Glaese%2C+A%2E%2C+McAleese%2C+N%2E%2C+and+Irving%2C+G%2E+Red+teaming+language+models+with+language+models%2E+arXiv+preprint+arXiv%3A2202%2E03286%2C+2022a%2E) [our card](https://md.research.elamur.ai/free/en/papers/perez-2022-red-teaming-language-models-language/)
34. Perez, E., Ringer, S., Lukošiūtė, K., Nguyen, K., Chen, E., Heiner, S., Pettit, C., Olsson, C., Kundu, S., Kadavath, S., et al. Discovering language model behaviors with model-written evaluations. arXiv preprint arXiv:2212.09251, 2022b. [find](https://scholar.google.com/scholar?q=Perez%2C+E%2E%2C+Ringer%2C+S%2E%2C+Lukos%CC%8Ciu%CC%84te%CC%87%2C+K%2E%2C+Nguyen%2C+K%2E%2C+Chen%2C+E%2E%2C+Heiner%2C+S%2E%2C+Pettit%2C+C%2E%2C+Olsson%2C+C%2E%2C+Kundu%2C+S%2E%2C+Kadavath%2C+S%2E%2C+et+al%2E+Discovering+language+model+behaviors+with+model%2Dwritten+evaluations%2E+) [our card](https://md.research.elamur.ai/free/en/papers/perez-2022-discovering-language-model-behaviors-model/)
35. PewResearch. Writing Survey Questions. u-s-surveys/writing-survey-questions/. Sap, M., Swayamdipta, S., Vianna, L., Zhou, X., Choi, Y., and Smith, N. A. Annotators with attitudes: How annotator beliefs and identities bias toxic language detection. In Association for Computational Linguistics (ACL),… [find](https://scholar.google.com/scholar?q=PewResearch%2E+Writing+Survey+Questions%2E+u%2Ds%2Dsurveys%2Fwriting%2Dsurvey%2Dquestions%2F%2E+Sap%2C+M%2E%2C+Swayamdipta%2C+S%2E%2C+Vianna%2C+L%2E%2C+Zhou%2C+X%2E%2C+Choi%2C+Y%2E%2C+and+Smith%2C+N%2E+A%2E+Annotators+with+attitudes%3A+How+annotator+belief) [our card](https://md.research.elamur.ai/free/en/papers/sap-2022-annotators-attitudes-how-annotator-beliefs/)
36. Saris, W. and Sniderman, P. Studies in public opinion: Attitudes, nonattitudes, measurement error, and change. Princeton University Press, 2004. [find](https://scholar.google.com/scholar?q=Saris%2C+W%2E+and+Sniderman%2C+P%2E+Studies+in+public+opinion%3A+Attitudes%2C+nonattitudes%2C+measurement+error%2C+and+change%2E+Princeton+University+Press%2C+2004%2E)
37. Simmons, G. Moral mimicry: Large language models produce moral rationalizations tailored to political identity. arXiv preprint arXiv:2209.12106, 2022. [find](https://scholar.google.com/scholar?q=Simmons%2C+G%2E+Moral+mimicry%3A+Large+language+models+produce+moral+rationalizations+tailored+to+political+identity%2E+arXiv+preprint+arXiv%3A2209%2E12106%2C+2022%2E)
38. Srivastava, A., Rastogi, A., Rao, A., Shoeb, A., Abid, A., Fisch, A., Brown, A., Santoro, A., Gupta, A., Garriga-Alonso, A., et al. Beyond the imitation game: Quantifying and extrapolating the capabilities of language models. arXiv preprint arXiv:2206.04615, 2022. [find](https://scholar.google.com/scholar?q=Srivastava%2C+A%2E%2C+Rastogi%2C+A%2E%2C+Rao%2C+A%2E%2C+Shoeb%2C+A%2E%2C+Abid%2C+A%2E%2C+Fisch%2C+A%2E%2C+Brown%2C+A%2E%2C+Santoro%2C+A%2E%2C+Gupta%2C+A%2E%2C+Garriga%2DAlonso%2C+A%2E%2C+et+al%2E+Beyond+the+imitation+game%3A+Quantifying+and+extrapolating+the+capabil) [our card](https://md.research.elamur.ai/free/en/papers/srivastava-2022-imitation-game-quantifying-extrapolating-capabilities/)
39. Uchendu, A., Ma, Z., Le, T., Zhang, R., and Lee, D. Turingbench: A benchmark environment for turing test in the age of neural text generation. arXiv preprint arXiv:2109.13296, 2021. [find](https://scholar.google.com/scholar?q=Uchendu%2C+A%2E%2C+Ma%2C+Z%2E%2C+Le%2C+T%2E%2C+Zhang%2C+R%2E%2C+and+Lee%2C+D%2E+Turingbench%3A+A+benchmark+environment+for+turing+test+in+the+age+of+neural+text+generation%2E+arXiv+preprint+arXiv%3A2109%2E13296%2C+2021%2E)
40. Xu, J., Ju, D., Li, M., Boureau, Y., Weston, J., and Dinan, E. Bot-adversarial dialogue for safe conversational agents. In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 2021. [find](https://scholar.google.com/scholar?q=Xu%2C+J%2E%2C+Ju%2C+D%2E%2C+Li%2C+M%2E%2C+Boureau%2C+Y%2E%2C+Weston%2C+J%2E%2C+and+Dinan%2C+E%2E+Bot%2Dadversarial+dialogue+for+safe+conversational+agents%2E+In+Proceedings+of+the+2021+Conference+of+the+North+American+Chapter+of+the+Assoc)
41. Yan, T. Consequences of asking sensitive questions in surveys. Annual Review of Statistics and Its Application, 2021. [https://www.pewresearch.org/our-methods/](https://www.pewresearch.org/our-methods/) [↗](https://www.pewresearch.org/our-methods/)