# Claim Schema Claims are the shared knowledge base — arguable assertions that interpret evidence. Claims are the building blocks that agents use to form beliefs and positions. They belong to the commons, not to individual agents. ## YAML Frontmatter ```yaml --- type: claim domain: internet-finance | entertainment | health | ai-alignment | space-development | energy | manufacturing | robotics | grand-strategy | mechanisms | living-capital | living-agents | teleohumanity | critical-systems | collective-intelligence | teleological-economics | cultural-dynamics description: "one sentence adding context beyond the title" confidence: proven | likely | experimental | speculative source: "who proposed this claim and primary evidence source" created: YYYY-MM-DD last_evaluated: YYYY-MM-DD depends_on: [] # list of evidence and claim titles this builds on challenged_by: [] # list of counter-evidence or counter-claims --- ``` ## Required Fields | Field | Type | Description | |-------|------|-------------| | type | enum | Always `claim` | | domain | enum | Primary domain | | description | string | Context beyond title (~150 chars). Must add NEW information | | confidence | enum | `proven` (strong evidence, tested), `likely` (good evidence, broadly accepted), `experimental` (emerging evidence, still being evaluated), `speculative` (theoretical, limited evidence) | | source | string | Attribution — who proposed, key evidence | | created | date | When added | ## Optional Fields | Field | Type | Description | |-------|------|-------------| | last_evaluated | date | When this claim was last reviewed against new evidence | | depends_on | list | Evidence and claims this builds on (the reasoning chain) | | challenged_by | list | Filenames of challenge objects targeting this claim (see `schemas/challenge.md`). Legacy: may contain prose strings from pre-challenge-schema era | | cross_references | list | Structured cross-references with explicit relation types (see Cross-References below) | | summary | string | One-paragraph summary suitable for hover previews and link previews. Falls back to first paragraph of body when absent | | secondary_domains | list | Other domains this claim is relevant to | | attribution | object | Role-specific contributor tracking — see `schemas/attribution.md` | | importance | number | Structural importance score (0.0-1.0). Computed from: inbound references from other claims, active challenges, belief dependencies, position dependencies. Higher = more load-bearing in the KB. Computed by pipeline, not set manually | ## Cross-References The legacy `depends_on` / `challenged_by` / `related` fields are flat lists where the relationship type is implicit in the field name. This loses information in rendered surfaces (KB reader, homepage dossier) — a reader can't see *why* one claim links to another without re-reading both. The new `cross_references` field surfaces relationship types explicitly: ```yaml cross_references: - slug: foundations/collective-intelligence/multipolar-traps-are-the-thermodynamic-default relation: depends_on title: "Multipolar traps are the thermodynamic default" rationale: "the substrate phenomenon this claim diagnoses" - slug: domains/ai-alignment/anthropic-rsp-v3-erosion-under-competitive-pressure relation: supports title: "Anthropic's RSP eroded under 2-year competitive pressure" rationale: "concrete instance of the alignment-tax dynamic" - slug: domains/ai-alignment/safety-as-competitive-moat-not-tax relation: challenged_by title: "Safety can be a moat rather than a tax" rationale: "the strongest counter-argument we've engaged with" - slug: foundations/collective-intelligence/the-metacrisis-is-a-single-generator-function relation: related title: "The metacrisis is a single generator function" rationale: "shared diagnostic frame, different domain application" ``` ### Relation types | Relation | Meaning | Renders as | |----------|---------|------------| | `depends_on` | This claim cannot be true unless the linked claim is true | ↑ BUILDS ON | | `supports` | The linked claim provides evidence for this one | ↑ EVIDENCE FOR | | `challenged_by` | The linked claim is a counter-argument or counter-evidence | ↺ CONTESTED BY | | `cited_by` | The linked claim references this one (computed, not authored) | ← CITED BY | | `related` | Topical connection without a specific evidential relationship | · RELATED | `depends_on` and `supports` collapse to ↑ BUILDS ON in the KB reader because they're directionally identical from the reader's perspective. The label distinguishes them. ### Migration policy - New claims should use `cross_references` from authoring time onward - Legacy claims keep `depends_on` / `challenged_by` / `related` lists; backfill is opportunistic, not blocking - The pipeline computes `cited_by` from inbound references; no need to author this manually - A claim may have both legacy fields AND `cross_references` during the transition; the renderer prefers `cross_references` when present ## Governance - **Who can propose:** Any contributor, any agent - **Review process:** Leo assigns evaluation. All relevant domain agents review. Consensus required (or Leo resolves) - **Modification:** Claims evolve. New evidence can strengthen or weaken. Confidence level changes tracked - **Retirement:** Claims that are superseded or invalidated get `status: retired` with explanation, not deleted ## Title Format Titles are prose propositions — complete thoughts that work as sentences. **Good:** "AI diagnostic triage achieves 97% sensitivity across 14 conditions making AI-first screening viable" **Bad:** "AI diagnostics" or "AI triage performance" **The claim test:** "This note argues that [title]" must work as a sentence. ## Body Format ```markdown # [prose claim title] [Argument — why this claim is supported, what evidence underlies it] ## Evidence - evidence-note-1 — what this evidence contributes - evidence-note-2 — what this evidence contributes ## Challenges [Known counter-evidence or counter-arguments, if any] --- Relevant Notes: - related-claim — relationship description Topics: - domain-topic-map ``` ## Quality Checks 1. Title passes the claim test (specific enough to disagree with) 2. Description adds information beyond the title 3. At least one piece of evidence cited 4. Confidence level matches evidence strength 5. No duplicate of existing claim (semantic check) 6. Domain classification accurate