What
Talk to the City (TttC) is a Large Language Model (LLM) based deliberation tool designed to significantly enhance collective decision-making processes. By capturing and organizing the diverse perspectives of populations ranging from 50 to over 5 million individuals, TttC aims to transform the way opinions are aggregated and understood on a massive scale. This tool generates interactive and comprehensive reports, such as summaries and chatbots representing specific viewpoints, offering a new paradigm for accessing and interpreting public discourse.
Who
The project is led by Bruno Marnette, with development and research facilitated by the team at Objective.is. The tool leverages OpenAI and Anthropic’s LLMs, which have been donated for this research endeavor.
Where & When
TttC is currently in the prototype testing phase.
Why
The primary motivation behind developing TttC is to address the current limitations in collective deliberation and decision-making processes. Traditional methods, such as categorical surveys, often fail to capture the depth and nuance of public opinion. TttC seeks to overcome these challenges by:
- Enhancing the quality of collective deliberation and coordination today.
- Enabling more sophisticated deliberation and representation mechanisms in an advanced AI-driven future.
- Facilitating better alignment and coordination among diverse groups and interests.
How
- Opinion Aggregation: Using LLMs to identify and cluster similar opinions, presenting them through a user-friendly visual interface.
- Simulated Deliberation: Developing LLM chat representatives for each opinion cluster, enabling simulated conversations that reveal the core of public discourse, disagreements, and potential consensus areas.
- Policy Iteration Support: Assisting policymakers in rapidly iterating on policy proposals by providing access to a synthesized representation of public opinion and simulated stakeholder reactions.