Every AI TownSquare session follows the same seven-phase protocol. That creates methodological comparability — the same method, in different rooms. But protocol alone doesn't create a dataset. The data becomes comparable when every session produces the same structured objects, relationships, and provenance. ATS-DATA-001 defines that common data layer.
"Schema" gets used loosely to mean several different things. Here's exactly what each artifact on this page is — and isn't.
The human-readable ontology this page renders directly: what each object means, how it relates to the others, why it exists.
schema.jsonMachine-enforceable structure — types, required fields, enums. Validates the shape of a dataset. Cannot check whether a cross-reference resolves.
ats-data-001-v1.1.0.schema.json →Checks what a JSON Schema structurally can't: that every claim_id, theme_id, and tension_id referenced by a relationship field actually exists elsewhere in that dataset.
scripts/validate-sessions.jsAn empty, conformant instance of the shape — the starting point for a new node to fill in. Not itself validation of anything.
aitownsquare-template-v1.1.0.json →A populated dataset that passes both validators — proof the specification can actually be instantiated, not just described.
CB-DXB-001 →A persistent AI TownSquare location — a city or institution that can host one or more Sessions over time. Distinct from a Session: Dubai is a Node; the 3 April 2026 Dubai dialogue is a Session at that Node.
This is why "show me everything that's happened in Dubai" or "compare Dubai's first session with its seventh" are answerable questions, not reverse-engineering exercises: node_id is a real field on every Session, not an inferred one.
That order is the order a dialogue unfolds in, inside one Session at one Node. It is not the order the data is built in. Underneath, every object references others by id, and not only forward: a Claim supports a Theme, or contradicts another Claim. A Theme shapes how a Signal reads. A Tension explains a gap in a specific SRI pillar and generates one or more Actions. At one session, that's a readable list. At a hundred, it's what makes the dataset queryable instead of just archived.
Specification source: schema.json. This page renders it directly. For the machine-enforceable version, see the JSON Schema; to start filling one in, download the v1.1.0 template →
A first-class property required on every claim, theme, tension, signal, and action — not optional metadata. It is what makes an object answerable to the question 'where did this come from?'
| Phase | What happens | Produces |
|---|---|---|
| Prime | Frames the session's central question with data and urgency. | Questions |
| Complicate* | Introduces nuance cards with competing evidence. | — |
| Position | The Catalyst models a principled stance to provoke disagreement; the first claims enter the record. | Claims |
| Breakout | Small-group deliberation on specific cases generates the bulk of raw claims. | Claims |
| Shareback | Claims are reported back to the room; quantitative polling mapped to SRI pillars produces signals. | Claims, Signals |
| Synthesize | The Catalyst extracts cross-cutting themes and names the tensions the room could not close. | Themes, Tensions |
| Capture | Actions are opened for the Readiness Institute; the brief package is confirmed for publication. | Actions, Brief |
*Complicate has no first-class output of its own — it feeds the Claims that Position and Breakout go on to produce.
A session at the Dubai node (DXB), and the first complete reference dataset built against ATS-DATA-001 v1.1.0 — every object cross-referenced, every object traceable to its source in the published Civic Brief. A new node in London, Berlin, or Singapore can produce conformant data against the same specification from its first session.