The published record we build on.
The mechanics of how opinion moves come from the published record, formula by formula. The behavioral axes that feed them are authorial, with adherence to the demographic profile measured — and we say which is which.
Foundations
Behavioral science & agent simulation.
01Daniel Kahneman. Thinking, Fast and Slow. Farrar, Straus & Giroux.2011
02Tversky & Kahneman. "Judgment under uncertainty: Heuristics and biases." Science 185.1974
03Robert Cialdini. Influence: The Psychology of Persuasion. Harper Business.1984
04Jonathan Haidt. The Righteous Mind. Pantheon Books.2012
05Damon Centola. How Behavior Spreads. Princeton University Press.2018
06Watts & Strogatz. "Collective dynamics of small-world networks." Nature 393.1998
07Park et al. "Generative Agents: Interactive Simulacra of Human Behavior." Stanford / arXiv.2023
08Thaler & Sunstein. Nudge. Yale University Press.2008
Frontier · Layer 01
Synthetic person — personas & individuals.
09"DeepPersona: A Generative Engine for Scaling Deep Synthetic Personas." Deep persona generation via a taxonomy of human attributes.2025
10"How Well Do LLMs Capture Human Personality?" On persona-manifold collapse in latent space.2026
11"The Need for a Socially-Grounded Persona Framework for User Simulation." Sociopsychological frameworks vs. isolated demographic attributes.2026
12"Eval4Sim: An Evaluation Framework for Persona Simulation." Adherence, consistency and naturalness against human corpora.2026
13"Beyond Cooperative Simulators: Realistic User Personas for Robust Evaluation." Evolutionary search with multi-objective fitness over personas.2026
14"Persona Vectors: Monitoring and Controlling Character Traits." Persona vectors in activation space.2025
Frontier · Layer 02
Synthetic society — emergent social simulation.
15Park et al. "Generative Agent Simulations of 1,000 People." Replicas of real individuals validated against the General Social Survey.2024
16"SocioVerse." A social-simulation world model over a pool of 10M real users.2025
17"AgentSociety: Large-Scale Simulation of LLM-Driven Agents." Computational experiments on polarization, basic income and shocks.2025
18"HumanStudy-Bench: AI Agent Design for Participant Simulation." Reconstructs human-subject studies and measures behavioral agreement.2026
19"Don't Trust Generative Agents — Unless You Benchmarked their Empirical Realism." Empirical realism must be validated, not assumed.2025
20"From ChatGPT to DeepSeek: Can LLMs Simulate Humanity?" A critical examination of the limits of simulating human behavior.2025
Frontier · Layer 03
Cognitive civilization — open-ended co-evolution.
21"Agentopia: Long-Term Life Simulation and Learning in Agent Societies." Agents develop social intelligence over simulated "years."2026
22"Static Sandboxes Are Inadequate: Open-Ended Co-Evolution in LLM Multi-Agent Simulations." Open-ended co-evolution vs. static sandboxes.2025
23"Cooperate or Collapse" (GovSim). Emergence of sustainable cooperation — or collapse — among agents.2024
Base
Grounding surveys.
24"From Individual to Society: A Survey on Social Simulation Driven by LLM-based Agents." Covers the individual → society progression.2024
Disclosure
Every estimate ships with declared uncertainty.
Bands, not points. Every behavioral estimate comes with its uncertainty band and calibration sources — never a bare number.
See how we measure →
Backtest updates
Get every new backtest by email.
When a population earns or loses a seal, or a new real outcome is measured against the engine, we write once. No newsletter cadence — only new entries in the ledger.