LRI defines non-operational invariants for protecting living identity in human-system relationships from optimization, capture, silent substitution, and continuity loss.

The central idea is that identity should not be treated as a fixed profile, a static score, or a target for optimization. Identity remains living only if systems preserve revisability, relational context, and the human's authority to continue becoming over time.

Many systems model humans as stable, optimizable objects. That creates recurring risks such as:

• identity freezing through static labeling
• silent transition from assistance to authorship
• memory persistence beyond consented scope
• optimization against hesitation, refusal, or ambiguity
• profiling that becomes destiny
• relational drift that is operationally useful but existentially harmful

These are not only product-design issues. They are failures of identity governance.

LRI provides a protocol and reference implementation for reasoning about identity continuity, authority, drift, and relational change without collapsing a living person into a fixed machine-readable essence.

The project currently includes:

• protocol-level identity and lifecycle schemas
• trust and security model docs
• a reference implementation with observer, drift, authority, security, and metrics services
• API and adapter skeletons
• playground scenarios and snapshots
• tests for continuity, metrics, observer behavior, DMP-lite integration, and security

A system can be operationally useful while still being identity-damaging.

Examples:

• a system infers who a person "really is" from past behavior
• a profile starts deciding treatment or opportunity
• a memory layer outlives consent boundaries
• an assistant becomes an author without an explicit boundary crossing
• drift is measured only for operational control, not for relational harm

LRI is useful because it treats these as governance and protocol problems, not only product bugs.

This repository already contains a concrete artifact, not only philosophical framing:

• protocol definitions in protocol/
• security and trust model documentation in docs/
• a Python reference implementation in lri-reference/
• playground scenarios and trajectory snapshots in playground/
• validation-facing assets and tests

Key files:

• docs/SECURITY_MODEL.md
• docs/architecture/lri-trust-model.md
• protocol/lri/schema/identity.yaml
• protocol/lri/schema/lifecycle.yaml
• lri-reference/services/authority_policy.py
• lri-reference/services/drift_monitor.py
• lri-reference/services/observer.py
• lri-reference/tests/
• VALIDATION_RESULTS.md

LRI is most relevant for failures such as:

• identity spoofing without continuity grounding
• history revisionism and loss of causal identity memory
• drift beyond declared trust boundaries
• optimization against agency, refusal, or revisability
• substitution of self-creation by automated system narratives

For broader framing, see docs/safety/identity_governance_threat_model.md.

The repository now exposes a simple root-level validation path:

• reference implementation tests
• deterministic project validation snapshot
• tracked validation results in VALIDATION_RESULTS.md

Run:

python scripts/validate_project.py
python scripts/generate_validation_results.py

Reference implementation tests:

cd lri-reference
python -m pytest -q

Explore the playground:

cd playground
python playground.py

LRI is not trying to optimize identity.

It is trying to define what must remain protected if a human identity is to stay living, revisable, and relational in the presence of systems that would otherwise compress it.