the_science_of_sentiment.

Arena is not just a chatbot. It is a high-fidelity social simulation grounded in demographic data, behavioral psychology, and the latest in large language models.

01_Demographic_Grounding

Every simulation begins with "Grounding". We ingest Census-level data (SA1/SA2 levels) to understand the specific makeup of a neighborhood. This includes age distribution, housing tenure (owners vs. renters), occupation categories, and language diversity.

Our synthesis engine then generates a "Community Roster"—a statistically representative set of virtual personas. If a neighborhood is 30% renters and 15% healthcare workers, the Arena simulation will reflect that exact ratio in its agent population.

02_The_Scoring_Matrix

Political Risk Index

A weighted composite score (0-100) that predicts the likelihood of project derailment. It factors in high-salience opposition, trust collapse rates, and the intensity of emotional responses from influential community cohorts.

Polarization Index

Measures the "gap" between community camps. By analyzing the sentiment variance between different demographic cohorts, we identify if a proposal is building consensus or driving a wedge into the social fabric.

Trust Collapse Rate

Specifically tracks the percentage of agents whose "Trust" metric falls below the critical threshold of 2.0. Once trust collapses, engagement shifts from "dialogue" to "resistance."

Inclusion Penalty

Measures the gap between the target demographic and the actual participation. If a simulation is dominated by a few vocal cohorts while minority groups remain silent, the Inclusion Penalty spikes.

03_Behavioral_Architecture

Semantic Appraisal

Agents don't just "read" text; they appraise it based on their values. A proposal for a new park is appraised differently by a parent (recreation value) than by a nearby homeowner (property value/noise concerns).

Emotional Inertia

Trust is hard to earn and easy to lose. Our agents maintain emotional memory. A single poor interaction can have long-lasting effects on their willingness to support future project phases.

our_operating_principles.

Statistical Integrity: Personas must match Census realities.

Transparency: No hidden prompts. Every agent 'thought' is logged.

Practitioner Focus: Simulation serves real-world outcomes.