---
title: "Analyzing Biases in Political Dialogue: Tagging U.S. Presidential"
description: "The authors argue that DAMSL-style dialogue act schemes describe conversational function well but lack explicit labels for… 24 references."
authors: ["Lavanya Prahallad", "Radhika Mamidi"]
year: 2025
venue: "arXiv:2505.19515"
source: "https://arxiv.org/abs/2505.19515"
doi: "10.48550/arxiv.2505.19515"
arxiv: "2505.19515"
licence: "CC BY 4.0 (arXiv)"
licence_url: "http://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/prahallad-2025-analyzing-biases-political-dialogue-tagging/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/prahallad-2025-analyzing-biases-political-dialogue-tagging/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2025](https://md.research.elamur.ai/free/en/papers/2025/page/16/) / Analyzing Biases in Political Dialogue: Tagging U.S. Presidential Debates with an Extended DAMSL Framework

# Analyzing Biases in Political Dialogue: Tagging U.S. Presidential Debates with an Extended DAMSL Framework

**Authors:** Lavanya Prahallad and [Radhika Mamidi](https://md.research.elamur.ai/free/en/authors/radhika-mamidi/) · **Year:** 2025 · **Venue:** arXiv:2505.19515 · **Licence:** [CC BY 4.0 (arXiv)](http://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://arxiv.org/abs/2505.19515) · [PDF](https://arxiv.org/pdf/2505.19515) · [DOI 10.48550/arxiv.2505.19515](https://doi.org/10.48550/arxiv.2505.19515) · [arXiv 2505.19515](https://arxiv.org/abs/2505.19515) · [OpenAlex](https://openalex.org/W4415035580)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
The authors argue that DAMSL-style dialogue act schemes describe conversational function well but lack explicit labels for rhetorical bias and confrontation in political talk. They propose BEADS (Bias-Enriched Annotation for Dialogue Structure), an extension of DAMSL. BEADS adds bias, emotional-appeal and adversarial labels, such as Political Bias, Appeals to Fear, Selective Emphasis, Rebuttal, Interruption and Turn Request, alongside standard act categories.

They applied it to the two 2024 U.S. presidential debates (Trump–Biden and Trump–Harris). The transcripts were checked against video, cleaned, and split into self-contained speech units. Each unit was then labeled in light of its surrounding turns. One trained expert produced the reference labels. ChatGPT-4o, prompted with Chain-of-Thought, labeled the same data and agreed with the expert about 70% of the time.

Tag counts show Trump leading in challenges, adversarial exchanges, fear appeals, political bias and dismissiveness in both debates. The authors present BEADS as a reusable, language-independent scheme for discourse analysis. However, the evidence comes from a single annotator and two debates.

## Key points
- **Corpus:** Trump–Biden (CNN, June 27, 2024) has 19,219 words and 1,450 sentences. Trump–Harris (ABC News, September 10, 2024) has 18,123 words and 1,472 sentences.
- **Background:** The original DAMSL had over 220 labels. Switchboard DAMSL cut this to 42 core tags. BEADS is described as adding 15 bias- and confrontation-oriented labels.
- **LLM vs. expert:** Zero-shot ChatGPT-4o with Chain-of-Thought matched the expert on about 70% of tags. The mismatches came mostly from misread context, overlapping rhetorical strategies, and subtle framing.
- **Example counts (Trump vs. opponent):**

  | Tag | vs. Biden | vs. Harris |
  |---|---|---|
  | Political Bias | 29 vs. 22 | 22 vs. 14 |
  | Appeals to Fear | 32 vs. 24 | 34 vs. 28 |
  | Adversarial Exchange | 17 vs. 9 | 13 vs. 6 |
  | Perceived Dismissiveness | 14 vs. 10 | 7 vs. 3 |

- **Context changes labels:** A bare "That's not true" reads as a Statement in isolation. In context it becomes a Challenge. A question asked after a vague proposal becomes an Adversarial Exchange.

## Interesting ideas
- The authors label each unit twice, once alone and once with neighbouring turns, and report which labels flip; the flip list shows how much an act depends on discourse
- Bias and confrontation tags sit as an additional layer over ordinary dialogue acts rather than replacing them, so a unit can be both an Answer and a Rebuttal
- The tag guide pairs each label with a definition, a generic invented example, and a real attested excerpt, which makes the scheme easier to teach to annotators or prompt to a model
- LLM–expert disagreements are sorted by cause (context misreading, overlapping strategies, subtle framing) instead of being reported only as a single accuracy figure
- Per-speaker tag counts across two comparable dialogues turn annotation into a profile of each participant's behaviour

## For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

- **I1** (frames): prag could run its speech-act classifier with and without the previous turn and log the flips, giving an explainable record of which acts are context-driven.
- **I2** (frames): speech-act frames in prag could carry a separate stance/rhetoric slot (challenge, fear appeal, dismissal) without disturbing the core act inventory.
- **I3** (learn): LLM teacher prompts for speech-act or frame labels could use this three-part card per label to reduce teacher confusion between neighbouring tags.
- **I4** (learn): when silver labels from an LLM teacher disagree with gold, a fixed error taxonomy could steer which cases feed repair rules and which are left as genuinely ambiguous.
- **I5** (world): the world of a dialogue text could keep per-speaker tallies of acts (challenges, interruptions, concessions) as speaker states, so answers about who dominated or who attacked whom can be traced to labelled turns.

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

## Institutions

- [International Institute of Information Technology, Hyderabad](https://md.research.elamur.ai/free/en/universities/international-institute-of-information-technology-hyderabad/) — University, Hyderabad, India · 38 works · 1,821 citations

📚 Shelves (5): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works · [🤔 Philosophy](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/philosophy/) · 284 works

🧭 Topics (1): [🗨️ Dialogue systems](https://md.research.elamur.ai/free/en/topics/dialogue-systems/) · 624 works

## Cite

Cite as: Lavanya Prahallad and Radhika Mamidi. 2025. Analyzing Biases in Political Dialogue: Tagging U.S. Presidential Debates with an Extended DAMSL Framework. arXiv:2505.19515.

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

```bibtex
@misc{prahallad2025analyzing,
  title = {{Analyzing Biases in Political Dialogue: Tagging U.S. Presidential Debates with an Extended DAMSL Framework}},
  author = {Lavanya Prahallad and Radhika Mamidi},
  year = {2025},
  doi = {10.48550/arxiv.2505.19515},
  eprint = {2505.19515},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2505.19515}
}
```

- [← previous An Information-Theoretic Approach to Identifying Formulaic Clusters in Textual Data](https://md.research.elamur.ai/free/en/papers/yoffe-2025-information-theoretic-approach-identifying-formulaic/)
- [next → Ancient and Rare Readings in the Syriac Harklean Version of the Gospel of John and their Significance for its Interpretation](https://md.research.elamur.ai/free/en/papers/jutkiewicz-2025-ancient-rare-readings-syriac-harklean/)
- [↑ back to papers of 2025](https://md.research.elamur.ai/free/en/papers/2025/page/16/)

## References

24 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Abdulaziz Aldayel and Walid Magdy. 2021. Unsupervised aspect-based stance detection in political debates. Journal of Computational Social Science, 4(2):401–417. [find](https://scholar.google.com/scholar?q=Abdulaziz+Aldayel+and+Walid+Magdy%2E+2021%2E+Unsupervised+aspect%2Dbased+stance+detection+in+political+debates%2E+Journal+of+Computational+Social+Science%2C+4%282%29%3A401%E2%80%93417%2E)
2. Harry Bunt, Jan Alexandersson, Jörg Carletta, Donna K. Byron, Kai-Yuh Hwang, Kiyong Lee, Laurent Romary, Mathew A. Purver, Manfred Stede, and David Traum. 2012. [Iso 24617-2: A semantically-based standard for dialogue annotation](https://md.research.elamur.ai/free/en/papers/bunt-2012-iso-24617-2-semantically-based/). LREC, pages 430– 437.
3. Harry Bunt, Katya Petukhova, and Alex Fang. 2018. The dialogbank: Dialogues with interoperable annotations. Language Resources and Evaluation, 52(3):645–674. [↗](https://doi.org/10.1007/s10579-018-9436-9)
4. CNN. 2024. Cnn presidential debate (june 27, 2024). [find](https://scholar.google.com/scholar?q=CNN%2E+2024%2E+Cnn+presidential+debate+%28june+27%2C+2024%29%2E)
5. Mark G. Core and James F. Allen. 1997a. Coding dialogs with the damsl annotation scheme. AAAI Fall Symposium on Communicative Action in Humans and Machines, pages 28–35. [find](https://scholar.google.com/scholar?q=Mark+G%2E+Core+and+James+F%2E+Allen%2E+1997a%2E+Coding+dialogs+with+the+damsl+annotation+scheme%2E+AAAI+Fall+Symposium+on+Communicative+Action+in+Humans+and+Machines%2C+pages+28%E2%80%9335%2E)
6. Mark G. Core and James F. Allen. 1997b. Coding dialogs with the damsl annotation scheme. Technical Report TR 98-42, University of Rochester. Available at University of Rochester Technical Report Archive. [find](https://scholar.google.com/scholar?q=Mark+G%2E+Core+and+James+F%2E+Allen%2E+1997b%2E+Coding+dialogs+with+the+damsl+annotation+scheme%2E+Technical+Report+TR+98%2D42%2C+University+of+Rochester%2E+Available+at+University+of+Rochester+Technical+Report+Archi)
7. Norman Fairclough. 2003. Analyzing Discourse: Textual Analysis for Social Research. Routledge. [find](https://scholar.google.com/scholar?q=Norman+Fairclough%2E+2003%2E+Analyzing+Discourse%3A+Textual+Analysis+for+Social+Research%2E+Routledge%2E)
8. Wei Hu and Qing Cao. 2023. Discourse strategies and ideological stance in chinese diplomatic speeches: A cda approach. Journal of Language and Politics, 22(1):1–22. [find](https://scholar.google.com/scholar?q=Wei+Hu+and+Qing+Cao%2E+2023%2E+Discourse+strategies+and+ideological+stance+in+chinese+diplomatic+speeches%3A+A+cda+approach%2E+Journal+of+Language+and+Politics%2C+22%281%29%3A1%E2%80%9322%2E)
9. Sajid Hussain and Muhammad Sajjad. 2022. Nlp-driven analysis of imran khan’s political speeches: A case study using ericksonian patterns. Digital Scholarship in the Humanities, 37(3):777–795. [find](https://scholar.google.com/scholar?q=Sajid+Hussain+and+Muhammad+Sajjad%2E+2022%2E+Nlp%2Ddriven+analysis+of+imran+khan%E2%80%99s+political+speeches%3A+A+case+study+using+ericksonian+patterns%2E+Digital+Scholarship+in+the+Humanities%2C+37%283%29%3A777%E2%80%93795%2E)
10. Cornelia Ilie. 2010. Identity work and ideological positioning in parliamentary discourse: Conflict between adversaries in uk parliamentary debates. Journal of Pragmatics, 42(4):885–901. [find](https://scholar.google.com/scholar?q=Cornelia+Ilie%2E+2010%2E+Identity+work+and+ideological+positioning+in+parliamentary+discourse%3A+Conflict+between+adversaries+in+uk+parliamentary+debates%2E+Journal+of+Pragmatics%2C+42%284%29%3A885%E2%80%93901%2E)
11. Daniel Jurafsky, Elizabeth Shriberg, and Debra Biasca. 1997. Switchboard dialog act corpus. In Switchboard Dialog Act Corpus. [find](https://scholar.google.com/scholar?q=Daniel+Jurafsky%2C+Elizabeth+Shriberg%2C+and+Debra+Biasca%2E+1997%2E+Switchboard+dialog+act+corpus%2E+In+Switchboard+Dialog+Act+Corpus%2E)
12. George Lakoff. 2004. Don’t Think of an Elephant!: Know Your Values and Frame the Debate. Chelsea Green Publishing. [find](https://scholar.google.com/scholar?q=George+Lakoff%2E+2004%2E+Don%E2%80%99t+Think+of+an+Elephant%21%3A+Know+Your+Values+and+Frame+the+Debate%2E+Chelsea+Green+Publishing%2E)
13. Angeliki Lazaridou, Angeliki Koufakou, and George Papadopoulos. 2020. A critical discourse analysis of trump’s campaign rhetoric using nlp. Discourse & Society, 31(5):497–514. [find](https://scholar.google.com/scholar?q=Angeliki+Lazaridou%2C+Angeliki+Koufakou%2C+and+George+Papadopoulos%2E+2020%2E+A+critical+discourse+analysis+of+trump%E2%80%99s+campaign+rhetoric+using+nlp%2E+Discourse+%26+Society%2C+31%285%29%3A497%E2%80%93514%2E)
14. ABC News. 2024. Abc news presidential debate (september 10, 2024). [find](https://scholar.google.com/scholar?q=ABC+News%2E+2024%2E+Abc+news+presidential+debate+%28september+10%2C+2024%29%2E)
15. Chikashi Nobata et al. 2023. [Dependency dialogue acts – annotation scheme and case study](https://md.research.elamur.ai/free/en/papers/cai-2023-dependency-dialogue-acts-annotation-scheme/). In Proceedings of the 2023 Workshop on Dialogue Structure.
16. Katya Petukhova and Ekaterina Kochmar. 2025. A fully automated pipeline for conversational discourse annotation: Tree scheme generation and labeling with large language models. arXiv preprint arXiv:2504.08961. [↗](https://arxiv.org/abs/2504.08961)
17. Teun A. van Dijk. 1997. Discourse as Social Interaction. SAGE Publications. [find](https://scholar.google.com/scholar?q=Teun+A%2E+van+Dijk%2E+1997%2E+Discourse+as+Social+Interaction%2E+SAGE+Publications%2E)
18. Marilyn A. Walker, Pranav Anand, Rob Abbott, and Ricky Grant. 2012. Stance classification using dialogic structure. In Proceedings of the 2012 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, pages 592–602. [find](https://scholar.google.com/scholar?q=Marilyn+A%2E+Walker%2C+Pranav+Anand%2C+Rob+Abbott%2C+and+Ricky+Grant%2E+2012%2E+Stance+classification+using+dialogic+structure%2E+In+Proceedings+of+the+2012+Joint+Conference+on+Empirical+Methods+in+Natural+Language)
19. Zeerak Waseem and Dirk Hovy. 2016. [Hateful symbols or hateful people? predictive features for hate speech detection on twitter](https://md.research.elamur.ai/free/en/papers/waseem-2016-hateful-symbols-or-hateful-people/). In Proceedings of NAACL-HLT, pages 88–93.
20. Hua Yu and Xiaoyu Zhao. 2021. A critical discourse analysis of biden’s 2021 state of the union address. Critical Discourse Studies, 18(4):367–384. [find](https://scholar.google.com/scholar?q=Hua+Yu+and+Xiaoyu+Zhao%2E+2021%2E+A+critical+discourse+analysis+of+biden%E2%80%99s+2021+state+of+the+union+address%2E+Critical+Discourse+Studies%2C+18%284%29%3A367%E2%80%93384%2E)
21. Sheng Yu and Lyle Ungar. 2021. Lexical and discourse-level framing in u.s. political speech. In Proceedings of the 2nd Workshop on Computational Approaches to Discourse, pages 111–121. [find](https://scholar.google.com/scholar?q=Sheng+Yu+and+Lyle+Ungar%2E+2021%2E+Lexical+and+discourse%2Dlevel+framing+in+u%2Es%2E+political+speech%2E+In+Proceedings+of+the+2nd+Workshop+on+Computational+Approaches+to+Discourse%2C+pages+111%E2%80%93121%2E)
22. Justine Zhang, Aron Culotta, and Cristian Danescu-Niculescu-Mizil. 2016. Conversational flow in oxford-style debates. In Proceedings of NAACL-HLT. [↗](https://doi.org/10.48550/arxiv.1604.03114)
23. Justine Zhang and Cristian Danescu-Niculescu-Mizil. 2016. Conversational flow in oxford-style debates. In Proceedings of NAACL-HLT, pages 1360–1370. [↗](https://doi.org/10.48550/arxiv.1604.03114)
24. Jieyu Zhao, Paul Resnick, and Qiaozhu Mei. 2022. Tracking political framing with neural discourse models. In Findings of the Association for Computational Linguistics: ACL 2022, pages 1254–1265. [find](https://scholar.google.com/scholar?q=Jieyu+Zhao%2C+Paul+Resnick%2C+and+Qiaozhu+Mei%2E+2022%2E+Tracking+political+framing+with+neural+discourse+models%2E+In+Findings+of+the+Association+for+Computational+Linguistics%3A+ACL+2022%2C+pages+1254%E2%80%931265%2E)

*Updated 1 November 2026*