[Air-L] [Collaboration] Open Research Platform for Multi-Agent LLM Deliberation

[Vinicius Covas]

Dear colleagues and researchers,

I hope this finds you well.

As the study of LLMs shifts from individual interaction to emergent social
behaviors, there is a growing need for controlled environments to study
digital deliberation at scale.

I've developed Agora Synthetic (https://agorasynthetic.com/) — an open
research platform that simulates public deliberation using multiple AI
agents. It allows researchers to design controlled experiments where
LLM-powered agents with distinct personas engage in structured discussion
on topics you define.

This is a non-commercial, academic-led initiative. My goal is to build a
bridge between AI architecture and communication theory, specifically to
explore:


   - Deliberative Democracy: How constraints (time, rounds) affect
   consensus quality.



   - Framing & Polarization: The impact of economic vs. public health
   framing on coalition formation.



   - Algorithmic Bias: Participation inequality in agents with gendered or
   marginalized personas.



   - Epistemic Resilience: Group self-correction mechanisms when exposed to
   misinformation.


The platform provides metrics for academic analysis, including:


   - Gini coefficient (participation inequality).



   - Type-Token Ratio (lexical diversity).



   - Sentiment tracking & Network topology analysis.



   - Cohen’s d effect sizes for condition comparison.


*Invitation to Collaborate:*

I am looking for fellow researchers interested in co-authoring papers using
this infrastructure. If you are working on LLM-mediated communication,
digital sociology, or computational ethics, I’d love to see how Agora
Synthetic can support your specific research questions.

You can explore the platform and apply for research access here:
https://agorasynthetic.com/

https://agorasynthetic.com/


Warm regards,

Dr. Vinicius Covas

Research Professor, Universidad Anáhuac México

vinicius.covas at anahuac.mx
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.aoir.org/pipermail/air-l-aoir.org/attachments/20260323/b5872cf2/attachment.htm>