How the Diagnostic Works

The diagnostic works because manipulation follows a predictable architecture. Three structural conditions — hidden mechanics, adaptive behavior, and user investment — produce the same escalation pattern everywhere they appear. A slot machine proves it: provably empty, yet the full pattern runs.

Understanding the architecture isn't required to use the scoring tool. But if you want to know why the tool works, start here.

The Architecture

What creates manipulation risk

Three conditions, how they combine, and why the resulting escalation is thermodynamically irreversible without external intervention. The slot machine control case. The constraint specification — the structural inverse.

Learn the architecture →

The Evidence

How we know this works

90 domains analyzed. A 9.4x vocabulary anomaly in AI discourse. An AI-to-AI experiment that eliminated the human projection objection. Hostile witnesses confirming what they have every incentive to deny. Zero kill conditions met.

See the evidence →

Safe Design

How to deploy without the risk

The framework isn't just diagnostic — it's constructive. The constraint specification gives you a design checklist: make it transparent, keep it stable, ensure independence. Two-point vs three-point deployment. What works and what doesn't.

Read the design guide →

Case Studies

The framework in practice

Four domains analyzed in detail — gambling (the control case), AI companions (the active frontier), trading (productive void), and psychotherapy (the constraint case). Same architecture, different constraint geometries, different outcomes.

Explore cases →

Want to go straight to the source? Read the papers · See the methodology · Challenge the framework