ArgsBase helps people talk through questions with AI. Several AI participants join the exchange, while a moderator and live summaries keep the discussion clear and easy to follow.
Built for discussion, not just one-shot answers.
Good decisions often need more than one polished answer. Most chat interfaces still do not support an open discussion that people can follow and shape.
One answer stream dominates even when the issue is contested.
Users receive outputs, but the system does not actively structure reasoning.
Agreements, conflicts, and open questions remain mostly hidden.
Different perspectives and roles appear in one shared interaction space.
Turn-taking and role assignment keep the exchange structured.
Live summaries and argument maps make the discussion inspectable.
A simple flow that keeps the exchange easy to follow while preserving participation, comparison, and reflection.
The user brings a question, dilemma, or contested topic.
The moderator sets up deliberators with distinct functions or perspectives.
Agents argue and respond under moderated turn-taking.
The user redirects, questions, and evaluates the exchange.
The analyzer highlights agreements, disagreements, and open questions.
The output helps users compare reasons and inspect uncertainty.
Paper, demo, and video for the EACL 2026 system demonstration.
Frieso Turkstra, Sara Nabhani, and Khalid Al Khatib. 2026. ARGSBASE: A Multi-Agent Interface for Structured Human-AI Deliberation. In Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 3: System Demonstrations), pages 563-574, Rabat, Morocco. Association for Computational Linguistics.