Elamur Research⌕
☰
home / papers / 2026 / Actors, Frames and Arguments: A Multi-Decade Computational Analysis of Climate Discourse in Financial News using Large Language Models

Actors, Frames and Arguments: A Multi-Decade Computational Analysis of Climate Discourse in Financial News using Large Language Models

Paper · 2026 · Ruiran Su, Markus Leippold and Janet B. Pierrehumbert

Summary

The paper looks at how elite financial journalism has covered climate change over time. This area matters because such coverage can influence investment and policy, yet it has received far less study than general news. The authors built the Dow Jones Climate News Corpus from Dow Jones Newswire articles published between 2000 and 2023. They then drew a stratified sample that keeps the corpus's diversity while favouring argument-rich texts. On this sample they ran an LLM-based Actor–Frame–Argument (AFA) pipeline that records three things: who speaks and with what stance, which frame dominates, and what claims and premises support each position.

Reliability is checked against a human-annotated gold standard. A Decompositional Verification Framework (DVF) also uses several judges to score each extraction on completeness, faithfulness, coherence and climate relevance. The longitudinal results show a structural change. Earlier coverage stressed risk, cost and regulatory burden. After the Paris Agreement, the emphasis moved to opportunity and innovation, and financial institutions became the leading voices. The work offers both an auditable, reusable method for long-horizon discourse analysis and evidence of how financial elites reframed the climate issue.

Key points

  • Corpus construction: The corpus holds 980,061 unique articles from 2000–2023. They were selected with climate-related Dow Jones Intelligent Identifiers (DJIDs), which a validation study estimated at about 95% precision and about 92% recall. Locality-sensitive hashing removed roughly 8.7% of articles as near-duplicates.
  • Sampling: A four-stage hierarchical procedure produced a 4,143-article analysis sample, about 0.8% of every period.
    • Articles were stratified into four periods: pre-crisis, financial crisis, post-crisis and climate finance surge.
    • Within each period, articles were clustered by theme using SBERT all-MiniLM-L6-v2 embeddings, Ward agglomerative clustering and silhouette-based selection of the number of clusters.
    • Maximal Marginal Relevance then picked representative articles from each cluster.
  • Extraction scheme: Actors are typed as company, financial institution, government, NGO/advocacy or individual. Each actor gets a stance (supportive, opposing, neutral or mixed), backed by an evidence quote from the article. Frames come from an eight-frame typology adapted from climate communication research.
  • Validation: Extractions were checked against a 2,000-article gold standard annotated by Zooniverse volunteers, together with the multi-judge DVF scoring.
  • Argument findings: Opportunity-oriented arguments carry more premises per claim (median 2.8–3.2) than risk arguments (1.4–1.7). Actor groups also argue differently: NGOs stress urgency, governments stress compliance, and firms and financial institutions present themselves as leaders of the transition.

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, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Arts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksSocial Sciences · 1,772 worksSociology and Political Science · 577 worksPhilosophy · 284 worksGeneral Social Sciences · 227 works

🧭 Topics 1Research subjects and languages the work is about.Discourse analysis · 466 works

Cite

Cite as Ruiran Su, Markus Leippold and Janet B. Pierrehumbert. 2026. Actors, Frames and Arguments: A Multi-Decade Computational Analysis of Climate Discourse in Financial News using Large Language Models. Findings of the Association for Computational Linguistics: EACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.findings-eacl.104

@inproceedings{su2026actors,
  title = {{Actors, Frames and Arguments: A Multi-Decade Computational Analysis of Climate Discourse in Financial News using Large Language Models}},
  author = {Ruiran Su, Markus Leippold and Janet B. Pierrehumbert},
  booktitle = {Findings of the Association for Computational Linguistics: EACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-eacl.104},
  url = {https://doi.org/10.18653/v1/2026.findings-eacl.104}
}

References

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

  1. Anthropic. 2024. Claude sonnet 4. Model documentation / system card. Claude Sonnet 4, proprietary model. find
  2. David M. Blei, Andrew Y. Ng, and Michael I. Jordan. 2003. Latent dirichlet allocation. Journal of Machine Learning Research, 3:993–1022.
  3. Maxwell T. Boykoff and Jules M. Boykoff. 2004. Balance as bias: Global warming and the u.s. prestige press. Global Environmental Change, 14(2):125– 136. find
  4. Jaime Carbonell and Jade Goldstein. 1998. The use of mmr, diversity-based reranking for reordering documents and producing summaries. In Proceedings of the 21st Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 335–336, Melbourne, Australia. find
  5. Dallas Card, Amber E. Boydstun, Justin H. Gross, Philip Resnik, and Noah A. Smith. 2015. The media frames corpus: Annotations of frames across issues. In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics (ACL) & 7th International Joint Conference on Natural Lang…
  6. Kaiping Chen, Ruitong Duan, Ying Peng, and Jingwen Zhang. 2023. How gpt-3 responds to different publics on climate change and black lives matter: A critical appraisal of equity in conversational ai. arXiv preprint arXiv:2209.13627. Preprint, last revised March 2023. — [arXiv:2209.13627](https://arxiv.org/abs/2209.13627) find
  7. Zheng Chen, Yulan He, and Yu Zhang. 2021. Hitting your marq: Multimodal argument quality assessment. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 6958–6970. Association for Computational Linguistics. find
  8. Gheorghe Comanici, Eric Bieber, Mike Schaekermann, Ice Pasupat, Noveen Sachdeva, Inderjit Dhillon, and Marcel Blistein et al. 2025. Gemini 2.5: Pushing the frontier with advanced reasoning, multimodality, long context, and next generation agentic capabilities. Technical report, Google / DeepMind. Pr…
  9. Dorottya Demszky, Nikhil Garg, Rob Voigt, James Zou, Jesse Shapiro, Matthew Gentzkow, and Dan Jurafsky. 2019. Analyzing polarization in social media: Method and application to tweets on 21 mass shootings. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Comp… find
  10. Dow Jones. 2024. Dow Jones Intelligent Identifiers (DJID) Taxonomy API. Proprietary subject classification system for financial news content. find
  11. Ran Duan and Serena Miller. 2021. Climate change in china: A study of news diversity in party-sponsored and market-oriented newspapers. Journalism, 22(10):2493–2510. find
  12. Robert M. Entman. 1993. Framing: Toward clarification of a fractured paradigm. Journal of Communication, 43(4):51–58. large language models estimate public opinion about global warming? an empirical assessment of algorithmic fidelity and bias. PLOS Climate, 3(8):e0000429. find
  13. Martin Gleize, Eyal Shnarch, Ran Levy, Ben Bogin, Ranit Aharonov, and Noam Slonim. 2020. Efficient pairwise annotation of argument quality. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (ACL), pages 5772–5781. Association for Computational Linguistics. find
  14. Markus Leippold. 2023. Thus spoke gpt-3: Interviewing a large-language model on climate finance. Finance Research Letters, 53:103617. find
  15. Deniz Gorur, Antonio Rago, and Francesca Toni. 2025. Can large language models perform relation-based argument mining? In Proceedings of the 31st International Conference on Computational Linguistics, pages 8518–8534, Abu Dhabi, UAE. Association for Computational Linguistics. find
  16. Ivan Habernal, Henning Wachsmuth, Iryna Gurevych, and Benno Stein. 2018. The argument reasoning comprehension task: Identification and reconstruction of implicit warrants. In Proceedings of NAACL-HLT 2018, pages 1930–1940.
  17. Valentin Hofmann, Xiaowen Dong, Janet Pierrehumbert, and Hinrich Schuetze. 2022. Modeling ideological salience and framing in polarized online groups with graph neural networks and structured sparsity. In Findings of the Association for Computational Linguistics: NAACL 2022, pages 536–550, Seattle,… find
  18. Xinyu Hua, Zichao Yang, and Yiming Yang. 2022. Iam: A comprehensive and large-scale dataset for integrated argument mining. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 2052– 2064. Association for Computational Linguistics. find
  19. Tunazzina Islam and Dan Goldwasser. 2024. Post-hoc study of climate microtargeting on social media ads with llms: Thematic insights and fairness evaluation. arXiv preprint arXiv:2410.05401. Preprint. find
  20. Hao Li, Viktor Schlegel, Yizheng Sun, Riza Batista-Navarro, and Goran Nenadić. 2025. Large language models in argument mining: A survey. find
  21. Yang Liu, Christian Stab, and Iryna Gurevych. 2021. A neural transition-based model for argumentation mining. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics (ACL), pages 2135–2149. Association for Computational Linguistics. find
  22. Yang Liu, Sheng Zhang, Jingbo Shang, and Jiawei Han. 2020. A top-down neural architecture towards text-level parsing of discourse relations. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (ACL), pages 6386–6395. Association for Computational Linguistics. find
  23. MetaAI. 2025. LLaMA 4: Maverick (17b) — a multimodal mixture-of-experts model. Model card / documentation for LLaMA-4 Maverick-17B (128 experts), accessed via ModelScope / model repositories. find
  24. Mistral. 2024. Mixtral 8×22b. https://huggingface.co/mistralai/Mixtral-8x22B-v0.1. Open-weight mixture-of-experts language model. ↗
  25. Matthew C. Nisbet. 2009. Communicating climate change: Why frames matter for public engagement. Environment: Science and Policy for Sustainable Development, 51(2):12–23. find
  26. OpenAI. 2024. Gpt-4o system card. arXiv preprint arXiv:2410.21276. Multimodal “omni” model by OpenAI.
  27. Ayush Jain, Manikandan Padmanaban, Jagabondhu Hazra, Shantanu Godbole, and Hendrik Hamann. 2024. Empowering sustainable finance: Leveraging large language models for climate-aware investments. In ICLR 2024 Workshop on Tackling Climate Change with Machine Learning. Climate Change AI. find
  28. David Patterson, Joseph Gonzalez, Urs Hölzle, Quoc Le, Chen Liang, Lluis-Miquel Munguia, Daniel Rothchild, David So, Maud Texier, and Jeff Dean. 2022. The carbon footprint of machine learning training will plateau, then shrink. Computer, 55(7):18– 28. find
  29. Klaus Krippendorff. 2004. Content Analysis: An Introduction to Its Methodology. Sage Publications. find
  30. Maja Pavlovic and Massimo Poesio. 2024. The effectiveness of llms as annotators: A comparative overview and empirical analysis of direct representation. In NLPerspectives 2024. Association for Computational Linguistics.
  31. J. Richard Landis and Gary G. Koch. 1977. The measurement of observer agreement for categorical data. Biometrics, 33(1):159–174. find
  32. John Lawrence and Chris Reed. 2019. Argument mining: A survey. Computational Linguistics, 45(4):765– 818.
  33. Sanguk Lee, Tai-Quan Peng, Matthew H. Goldberg, Seth A. Rosenthal, John E. Kotcher, Edward W. Maibach, and Anthony Leiserowitz. 2024. find
  34. Can Rashmi Prasad, Bonnie Webber, Alan Lee, and Aravind Joshi. 2019. Penn discourse treebank 3.0. Linguistic Data Consortium. find
  35. ProQuest. 2024. Home - Dow Jones Factiva -LibGuides. https://proquest.libguides.com/ factiva. Describes DJID as containing approximately 350,000 classification codes. ↗
  36. Qwen. 2025. Qwen3 30b a3b. https://huggingface.co/unsloth/Qwen3-30B-A3B-GGUF. Open-weight evaluator model. ↗
  37. Nils Reimers and Iryna Gurevych. 2019. Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 3982–3992,…
  38. Margaret E. Roberts, Brandon M. Stewart, Dustin Tingley, Christopher Lucas, Jetson Leder-Luis, Sharon Gadarian, Bethany Albertson, and David Rand. 2014. Structural topic models for open-ended survey responses. American Journal of Political Science, 58(4):1064–1082. find
  39. Peter J. Rousseeuw. 1987. Silhouettes: a graphical aid to the interpretation and validation of cluster analysis. Journal of Computational and Applied Mathematics, 20:53–65. find
  40. Maria Ruiz-Dolz, Santiago Cortes, Javier Garcia, and Ana Garcia-Serrano. 2021. Transformer-based models for automatic identification of argument relations. IEEE Intelligent Systems, 36(4):54–62. find
  41. Tatjana Schirmag, Jakob H. Wedemeyer, Annika Stechemesser, and Leonie Wenz. 2025. Neural topic modeling reveals german television’s climate change coverage. Communications Earth & Environment, 6(1):441. find
  42. Hannah Schmid-Petri, Silke Adam, Ivo Schmucki, and Thomas Häussler. 2017a. A changing climate of skepticism: The factors shaping climate change coverage in the us press. Public Understanding of Science, 26(4):498–513. find
  43. Christian Stab and Iryna Gurevych. 2017a. Parsing argumentation structures in persuasive essays. In Proceedings of the 2017 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 985–995, Los Angeles, California. Association for…
  44. Christian Stab and Iryna Gurevych. 2017b. Parsing argumentation structures in persuasive essays. In Computational Linguistics, volume 43, pages 619– 659.
  45. Zhen Tan, Dawei Li, Song Wang, Alimohammad Beigi, Bohan Jiang, Amrita Bhattacharjee, Mansooreh Karami, Jundong Li, Lu Cheng, and Huan Liu. 2024. Large language models for data annotation and synthesis: A survey. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processin…
  46. Assaf Toledo, Yoav Kantor, Gabriel Stanovsky, and Ido Dagan. 2021. Graph-based argument quality assessment. In Proceedings of the International Conference on Recent Advances in Natural Language Processing (RANLP 2021), pages 1427–1437. find
  47. Petter Törnberg. 2024. Best practices for text annotation with large language models. arXiv preprint arXiv:2402.05129. — [arXiv:2402.05129](https://arxiv.org/abs/2402.05129) find
  48. Joe H. Ward. 1963. Hierarchical grouping to optimize an objective function. Journal of the American Statistical Association, 58(301):236–244. find
  49. Nicolas Webersinke, Mathias Kraus, Julia Anna Bingler, and Markus Leippold. 2022. Climatebert: A pretrained language model for climate-related text. In Proceedings of the AAAI 2022 Fall Symposium: The Role of AI in Responding to Climate Challenges. ArXiv:2110.12010. — [arXiv:2110.12010](https://arxiv.org/abs/2110.12010) find
  50. Hannah Schmid-Petri, Silke Adam, Ivo Schmucki, and Thomas Häussler. 2017b. A changing climate of skepticism: The factors shaping climate change coverage in the u.s. press. Public Understanding of Science, 26(4):498–513. find
  51. Qilong Wu, Thibault Masson, Aritra Mukherjee, Yihao Zhao, Chaima Driouich, Xingjian Xing, Patrick Paroubek, and Mickael Coustaty. 2024. Susgen-gpt: A data-centric llm for financial nlp and sustainability report generation. arXiv preprint arXiv:2412.10906. Preprint. find
  52. Hannah Schmid-Petri and Dorothee Arlt. 2016a. Constructing an illusion of scientific uncertainty? framing climate change in german and british print media. Communications, 41(3):265–289. find
  53. Liu Zhuang, Lin Wayne, Shi Ya, and Zhao Jun. 2021. A robustly optimized BERT pre-training approach with post-training. In Proceedings of the 20th Chinese National Conference on Computational Linguistics, pages 1218–1227, Huhhot, China. Chinese Information Processing Society of China.
  54. Hannah Schmid-Petri and Dorothee Arlt. 2016b. Constructing an illusion of scientific uncertainty? framing climate change in german and british print media. Communications, 41(3):265–289. find
  55. Robert J. Shiller. 2017. Narrative economics. American Economic Review, 107(4):967–1004. find
  56. Christian Stab and Iryna Gurevych. 2014. Annotating argument components and relations in persuasive essays. In Proceedings of COLING 2014, pages 1501– 1510.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/su-2026-actors-frames-arguments-multi-decade/