New LLM pipeline models civic deliberation with action-aware personas

Researchers have developed a novel pipeline to create speaker-attributed transcripts from public civic deliberation recordings, such as court hearings and school board meetings. This pipeline enriches transcripts with persona profiles, topics, and specific action tags like '[propose_motion]'. By fine-tuning LLM personas on this action-aware data, simulations of civic discourse show significant improvements in persona fidelity, consistency, and institutional fidelity, with simulated excerpts becoming difficult to distinguish from real deliberations. AI

IMPACT Enables more realistic and data-grounded simulations of civic discourse, potentially improving policy analysis and public engagement tools.

RANK_REASON The cluster describes a new research paper detailing a novel pipeline and datasets for LLM persona modeling in civic deliberation. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• arXiv
• LLM
• Point of Order: Action-Aware LLM Persona Modeling for Data-Grounded Civic Deliberation

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →