[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-81570-en":3,"doc-seo-81570-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},81570,549758146520,"Patrick","https://ap-avatar.wpscdn.com/avatar/80002397d8c0411e94?_k=1775819394049821470",8,"Research & Report","Point of Order Action-Aware LLM Persona Modeling for Data-Grounded Civic Deliberation","LLM-based simulations can study civic deliberation with control, yet existing approaches lack speaker-attributed data and evaluation methods for long-form institutional behavior. A reproducible pipeline converts public Zoom recordings into speaker-attributed transcripts enriched with persona profiles, topics, and pragmatic action tags such as [propose_motion]. Three datasets are released (Appellate Court, School Board, Municipal Council) and LLM personas are fine-tuned on the action-aware supervision. Evaluation across persona fidelity, persona consistency, institutional fidelity, and behavioral coherence shows strong gains, including up to 67% lower perplexity and up to 70% improved deliberative responsiveness.","Point of Order: Action-Aware LLM Persona Modeling for Data-Grounded  \nCivic Deliberation  \nScott Merrill Shashank Srivastava  \nUniversity of North Carolina at Chapel Hill  \n{smerrill, [ssrivastava}@cs.unc.edu](ssrivastava}@cs.unc.edu)  \narXiv :2511 . 178 13v 3 [ cs .CL] 10 Jul 2026  \nAbstract  \nLLM-based simulations can enable controlled studies of civic deliberation, but current systems lack speaker-attributed data and methods for evaluating long-form institutional behavior. ASR transcripts typically use anonymous labels such as Speaker_ 1, preventing models from learning stable participant behavior across meetings. We present a reproducible pipeline that converts public Zoom recordings into speaker-attributed transcripts enriched with persona profiles, topics, and pragmatic action tags such as [propose_motion] . Using this pipeline, we release three public datasets of government deliberation (Appellate Court hearings, School Board meetings, and Municipal Council sessions) and fine-tune LLM personas on this action-aware data. We evaluate simulations along four dimensions: persona fidelity, persona consistency, institutional fidelity, and behavioral coherence. Action-aware fine-tuning cuts perplexity by 67%, doubles classifier-based persona fidelity, increases vote attempts by up to 3.6×, and improves deliberative responsiveness by up to 70% . Human evaluations show that simulated excerpts are often hard to distinguish from real deliberations, indicating a practical foundation for data-grounded civic simulation studies.  \n1 Introduction  \nDeliberative forums such as courtrooms, city councils, and school boards shape high-stakes decisions in law, infrastructure, and education. These settings are not just generic multi-party conversations: they involve recurring participants, formal roles, procedural constraints, and decision points. LLMbased simulations could support controlled “whatif” analyses of such settings. For example, how deliberations change under different agendas, participation rules, or procedural structures. However, building such simulations requires models that can represent stable participants and institution-specific behavior over extended interactions.  \nFigure 1: Public Zoom recordings are converted into speaker-attributed transcripts and linked across videos to build large-scale datasets. Structured topics, action tags, and speaker profiles condition PEFT fine-tuned LLM personas, which are evaluated across four dimensions and validated through human judgments.  \nA central bottleneck is data. Although millions of public meeting recordings are available online, ASR-generated transcripts typically use anonymous labels such as Speaker_ 1 and Speaker_ 2, preventing models from learning consistent participant behavior across meetings. We address this with a lightweight multimodal speaker-linking pipeline (Figure 1, top) that uses visual cues, audio, and textual context to assign stable speaker identities in standard public Zoom recordings, without requiring specialized metadata.  \nSpeaker identity alone, however, is insufficient for realistic institutional simulation. Civic meetings move rapidly between formal procedure, spon-  \ntaneous discussion, questions, motions, clarifications, and votes. To model this structure, we introduce action-aware persona modeling (Figure 1, middle): LLM personas are trained on speakerattributed transcripts enriched with persona profiles, meeting topics, and pragmatic action tags such as [propose_motion] and [ask_clarification] . This supervision allows models to learn not only how individual speakers sound, but also what kinds of actions they take in institutional context.  \nUsing this pipeline, we release three speakerlabeled datasets of government deliberation (Appellate Court hearings, School Board meetings, and Municipal Council sessions) covering 1.72M words. We also release the speaker-linking and annotation pipeline as a reusable artifact for constructing similar datasets fro","cbCaieSlnRSCXyDt","https://ap.wps.com/l/cbCaieSlnRSCXyDt","pdf",4855666,4,1,39,"English","en",105,"# Introduction\n# Related Work\n## Speaker Diarization and Linking","[{\"question\":\"What problem does the paper address in LLM simulations of civic deliberation?\",\"answer\":\"Current systems struggle because ASR transcripts use anonymous speaker labels, which prevents models from learning stable participant behavior across meetings, and existing evaluations do not cover long-form institutional dynamics well.\"},{\"question\":\"How does the proposed pipeline build speaker-attributed transcripts from Zoom recordings?\",\"answer\":\"It uses a lightweight multimodal speaker-linking approach that leverages visual cues, audio, and textual context to assign stable speaker identities without requiring specialized metadata, then enriches transcripts with persona profiles and structured action tags.\"},{\"question\":\"What improvements does action-aware fine-tuning deliver in the experiments?\",\"answer\":\"Action-aware fine-tuning reduces perplexity by 67%, roughly doubles classifier-based persona fidelity, increases vote attempt rates by up to 3.6×, and improves deliberative responsiveness by up to 70%, with human evaluations finding excerpts often hard to distinguish from real transcripts.\"}]",1784174382,98,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"point-of-order-action-aware-llm-persona-modeling-for-data-grounded-civic-deliberation","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":20},"https://docshare.wps.com/document/point-of-order-action-aware-llm-persona-modeling-for-data-grounded-civic-deliberation/81570/",{"url":52,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-25","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does the paper address in LLM simulations of civic deliberation?","Question",{"text":75,"@type":76},"Current systems struggle because ASR transcripts use anonymous speaker labels, which prevents models from learning stable participant behavior across meetings, and existing evaluations do not cover long-form institutional dynamics well.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the proposed pipeline build speaker-attributed transcripts from Zoom recordings?",{"text":80,"@type":76},"It uses a lightweight multimodal speaker-linking approach that leverages visual cues, audio, and textual context to assign stable speaker identities without requiring specialized metadata, then enriches transcripts with persona profiles and structured action tags.",{"name":82,"@type":73,"acceptedAnswer":83},"What improvements does action-aware fine-tuning deliver in the experiments?",{"text":84,"@type":76},"Action-aware fine-tuning reduces perplexity by 67%, roughly doubles classifier-based persona fidelity, increases vote attempt rates by up to 3.6×, and improves deliberative responsiveness by up to 70%, with human evaluations finding excerpts often hard to distinguish from real transcripts.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]