Elamur Research⌕
☰
home / papers / 2026 / ARGSBASE: A Multi-Agent Interface for Structured Human–AI Deliberation

ARGSBASE: A Multi-Agent Interface for Structured Human–AI Deliberation

Paper · 2026 · Frieso Turkstra, Sara Nabhani and Khalid Al Khatib

Summary

Few tools help people deliberate, meaning weigh arguments on complex or contested issues. Existing ones, such as Discussion Tracker and BCause.app, make little use of LLM agents. The authors, from the University of Groningen, present ArgsBase, a web-based system in which a user discusses a self-chosen topic with several LLMs acting as fellow deliberators.

A moderator agent sets the agenda, decides who speaks next, asks for clarification of vague claims and keeps the exchange on topic, loosely following pragma-dialectics. A separate analyzer agent works in the background. It continuously reports a summary, the points of agreement, unresolved questions and an argument map.

The models are framed as partners for reasoning rather than as answer machines, in line with the "reasonable parrots" proposal of Musi et al. (2025). In a user study, participants found the interface easy to use and helpful for seeing other viewpoints. They asked for a clearer argument map and a less formal tone. The system offers an accessible testbed for research on hybrid human–AI argumentation.

Key points

  • Architecture: ArgsBase has four roles: a Moderator, LLM Deliberators, the human user (counted as the third deliberator), and a non-speaking Analyzer.
  • Turn-taking rules: No agent may speak twice in a row. The user must take part at least once every three turns, and the input box is outlined in blue when it is the user's turn.
  • Analyzer side panel: It has four sections: Conversation Summary, Points of Agreement, Open Questions, and an Argument Map of claims and their premises.
  • Deployment: The system runs on AWS with serverless Lambda functions that orchestrate the agents. A public demo is available at argsbase.chat, with a recorded walkthrough, as part of the EACL 2026 System Demonstrations.
  • Limitations and future work:
    • Observed problems include sycophancy or token disagreement, hard-to-control reply length, and agents sidestepping fallacies instead of explicitly labelling them.
    • Planned additions are a researcher configuration panel, downloadable logs, and a multi-participant voice version.

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

🎓University of GroningenGroningen · Netherlands70 works · 5,142 citations

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksSocial Sciences · 1,772 worksSafety Research · 543 works

Cite

Cite as Frieso Turkstra, Sara Nabhani and Khalid Al Khatib. 2026. ARGSBASE: A Multi-Agent Interface for Structured Human–AI Deliberation. Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 3: System Demonstrations).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.eacl-demo.39

@inproceedings{turkstra2026argsbase,
  title = {{ARGSBASE: A Multi-Agent Interface for Structured Human–AI Deliberation}},
  author = {Frieso Turkstra, Sara Nabhani and Khalid {Al Khatib}},
  booktitle = {Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 3: System Demonstrations)},
  year = {2026},
  doi = {10.18653/v1/2026.eacl-demo.39},
  url = {https://doi.org/10.18653/v1/2026.eacl-demo.39}
}

References

17 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Sarah Howard, Jo Tondeur, Maarten De Laat, Simon Buckingham Shum, Dragan Gašević, and George Siemens. 2022. Rethinking the entwinement between artificial intelligence and human learning: What capabilities do learners need for a world with ai? Computers and Education: Artificial Intelligence, 3:1000… find
  2. Petter Bae Brandtzaeg, Marita Skjuve, and Asbjørn Følstad. 2022. My ai friend: How users of a social chatbot understand their human–ai friendship. Human Communication Research, 48(3):404–429. find
  3. Stephanie Burkhalter, John Gastil, and Todd Kelshaw. 2006. A conceptual definition and theoretical model of public deliberation in small face-to-face groups. Communication Theory, 12(4):398–422. find
  4. John S. Dryzek, André Bächtiger, Simone Chambers, Joshua Cohen, James N. Druckman, Andrea Felicetti, James S. Fishkin, David M. Farrell, Archon Fung, Amy Gutmann, Hélène Landemore, Jane Mansbridge, Sofie Marien, Michael A. Neblo, Simon Niemeyer, Maija Setälä, Rune Slothuus, Jane Suiter, Dennis Thomp… find
  5. Yilun Du, Shuang Li, Antonio Torralba, Joshua B. Tenenbaum, and Igor Mordatch. 2024. Improving factuality and reasoning in language models through multiagent debate. In Proceedings of the 41st International Conference on Machine Learning, ICML’24. JMLR.org.
  6. Frans H. van Eemeren and Rob Grootendorst. 2003. A Systematic Theory of Argumentation: The pragma-dialectical approach. Cambridge University Press. find
  7. Geoffrey Irving, Paul Christiano, and Dario Amodei. 2018. Ai safety via debate. Preprint, arXiv:1805.00899. find
  8. Takayuki ITO, Mikoto OKUMURA, Takanori ITO, and Eizo HIDESHIMA. 2015. Implementation of a large-scale discussion support system collagree. Journal of Japan Industrial Management Association, 66(2):83– 108. find
  9. Jinghui Jiang, Amanda J Karran, Constantinos K Cour-saris, Pierre-Majorique Léger, and Jörg Beringer. 2022. A situation awareness perspective on human-ai interaction: Tensions and opportunities. International Journal of Human-Computer Interaction, 39(9):1789–1806. find
  10. Elena Musi, Nadin Kökciyan, Khalid Al Khatib, Davide Ceolin, Emmanuelle Dietz, Klara Maximiliane Gutekunst, Annette Hautli-Janisz, Cristián Santibáñez, Jodi Schneider, Jonas Scholz, Cor Steging, Jacky Visser, and Henning Wachsmuth. 2025. Toward reasonable parrots: Why large language models should ar… find
  11. N Muthmainnah, PMI Seraj, and Ibrahim Oteir. 2022. Playing with ai to investigate human-computer interaction technology and improving critical thinking skills to pursue 21st century age. Education Research International, 2022:1–17. find
  12. Someen Park, Jaehoon Kim, Seungwan Jin, Sohyun Park, and Kyungsik Han. 2024. PREDICT: Multi-agent-based debate simulation for generalized hate speech detection. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 20963–20987, Miami, Florida, USA. Associat… find
  13. Ashish Sharma, I Wei Lin, Adam S Miner, David C Atkins, and Tim Althoff. 2023. Human–ai collaboration enables more empathic conversations in text-based peer-to-peer mental health support. Nature Machine Intelligence, 5(1):46–57. find
  14. Polyxeni Vassilakopoulou, Arve Haug, Lars M Salvesen, and Ilias O Pappas. 2022. Developing human/ai interactions for chat-based customer services: lessons learned from the norwegian government. European Journal of Information Systems, 32(1):10–22. find
  15. Priyanka Kargupta, Ishika Agarwal, Tal August, and Jiawei Han. 2025. Tree-of-debate: Multi-persona debate trees elicit critical thinking for scientific comparative analysis. Preprint, arXiv:2502.14767. find
  16. Tian Liang, Zhiwei He, Wenxiang Jiao, Xing Wang, Yan Wang, Rui Wang, Yujiu Yang, Shuming Shi, and Zhaopeng Tu. 2024. Encouraging divergent thinking in large language models through multi-agent debate. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 17…
  17. Lina Markauskaite, Rebecca Marrone, Oleksandra Poquet, Simon Knight, Roberto Martinez-Maldonado, find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/turkstra-2026-argsbase-multi-agent-interface-structured/