Belief

slug: belief element_type: TERM mutability: MUTABLE inline: true current_version: 1 contentURI: null

A belief is an atomic unit of a user's personal governance — a single value, principle, or rule the user holds. Beliefs are typed (e.g., concept, axiom, directive) and editable; users add, modify, version, and retire their own beliefs over time. A user's full belief set IS their personal constitution.

What this term defines

The smallest meaningful unit of personal governance. A belief is more than a journal entry (which is raw, narrative) — it's a structured assertion about how the user thinks/acts/values. Examples:

• A concept: @autonomy = "the right to define one's own boundaries"
• An axiom: #always_revocable = "any delegated authority must be revocable"
• A directive: !morning_reflection = "10 minutes journaling before email"

Properties of a belief

1. Typed — concept (definition), axiom (principle), directive (rule for action)
2. Versioned — beliefs evolve; old versions retained for time-travel reflection
3. Linked — beliefs reference each other (a directive !morning_reflection may reference axioms #start_intentional, #protect_focus)
4. Private by default — beliefs live in data_on_device storage; user chooses what (if anything) to expose via persona
5. Aggregated into persona — the sanitized persona is derived from the belief set

Why mutable

Belief typology (the exact set of types: concept/axiom/directive) and linkage rules may evolve as users + the protocol learn what works. Governance vote can refine the schema.

Related elements

• persona — derived sanitized representation
• data_on_device — storage residency
• identity_sovereignty — beliefs are user-authored, user-owned

Implementation notes

Anima (mirrorX) implements beliefs as Drift tables (beliefs, belief_versions, belief_edges). The on-device Gemma model interacts with beliefs via the NativeBeliefToolSet (27-tool catalog — see mirrorX/CLAUDE.md Wave 2). Cross-platform sync of beliefs (v3+) uses encrypted P2P per data_on_device rule.