Debate Tutorial

Learn the NeoEnc debate syntax

Interactive

A guided walkthrough of the exact markdown-shaped grammar used on the landing page, with move-by-move examples and interactive confidence metadata.

This page turns the landing-page grammar block into a usable tutorial. It starts with atomic claims, shows how inference and relation moves work, then lets you manipulate a probability distribution that can live inside claim metadata.

Back to landingOpen debate arena
Claim

Claim

An atomic statement with an ID, text, and optional reason source.

Inference

Inference

A move that connects premises to a conclusion with an explicit inference type.

Relation

Relation

Support or opposition edges that keep pressure and disagreement visible.

Distribution

Distribution

Confidence metadata that turns truth judgments into inspectable shapes instead of a bare scalar.

Section 1

Pick a move type

Switch between real serializer outputs. The code sample updates from generated markdown, not hand-written approximations.

Every example below is produced by `DebateTreeAdapterImpl.toMarkdown`, so the tutorial stays in sync with the data model.

Move selector

Start with one claim

Every thread begins with an atomic claim. Give it an ID, plain text, and optionally a reason source.

  • Claim lines use square-bracket IDs like `[1]:`.
  • The trailing reason tag records the source of the claim.
  • Keep each claim atomic enough that others can support or challenge it cleanly.

Generated Syntax

Serializer output

Read the syntax as a ledger of moves rather than as one flat document.

Section 2

Edit uncertainty metadata

Claims can carry a full 1D distribution under the `p` metadata field. Move the controls and watch both the plot and the generated snippet update.

Claim snippet with probability metadata

This is the shape embedded in a claim's frontmatter before the `[id]:` line.

Live Graph

Probability graph

The green curve is your current claim confidence distribution. The dashed comparison curve is the broader baseline.

Distribution

Claim confidence distribution

A lightweight client-side plot of the same 1D distribution shape used for claim confidence metadata.

Count
6
Mean
0.70
Variance
0.04
Range
0 to 1
Continuity
continuous

Section 3

Read a full thread

Thread metadataProbabilityRelations

This example mixes thread metadata, confidence metadata, claims, probabilistic inference, and a con relation.

What to notice

  • Thread metadata appears once at the top and scopes the whole exchange.
  • Claim-level `p` metadata stays near the claim it qualifies.
  • The probabilistic inference move references existing claims instead of retyping new prose.
  • The con relation lets an objection pressure the main claim without pretending to conclude it.

How to use this page

Start in the move selector, then drag the uncertainty controls, then read the full thread. By the end you should be able to recognize the grammar as a sequence of typed moves rather than a block of mysterious symbols.

Read docs next
--- date: 2026-01-10 location: tutorial-lab tags: ["demo","debate-grammar","probability"] --- 1. Move by alice at 2026-01-10T09:00:00.000Z --- p: {"count":18,"mean":0.56,"variance":0.03,"range":[0,1],"continuity":true} --- [1]: Public institutions can accumulate epistemic capital. [reason: reason] 2. Move by bob at 2026-01-10T09:01:00.000Z [2]: Long-lived archives preserve criticism and corrections. [reason: testimony] 3. Move by charlie at 2026-01-10T09:02:00.000Z [3]: Preserved criticism improves future judgment. [reason: reason] 4. Move by charlie at 2026-01-10T09:03:00.000Z // (2) Long-lived archives preserve criticism and corrections. // (3) Preserved criticism improves future judgment. // ---- // (1) Public institutions can accumulate epistemic capital. // probabilistic 5. Move by dana at 2026-01-10T09:04:00.000Z [5]: Institutions can also ossify and preserve bad norms. [reason: memory] 6. Move by dana at 2026-01-10T09:05:00.000Z (5) .con (1)