# Score Separability **Domain:** Algorithmic Governance / Rights / Administrative Systems **Doc Type:** Canonical Constitutional Safeguard **Maturity:** Developed ## Definition **Score separability** is the requirement that a classification created for one legitimate purpose must not silently determine unrelated rights, services or judgments of human worth. A climate-vulnerability score should not become a credit stigma. A health risk should not become an employment penalty. A work-reliability score should not gate food or medical care. A displacement credential should not become a security designation. Each domain requires its own data limits, decision rules, explanation, review and remedy. This is more than the mathematical separability of component scores. It is a constitutional firewall between domains of life. ## Climate and Reparative Context [[articles/Climate Justice and Global Reparative Systems|Climate Justice and Global Reparative Systems]] treats score separability as essential to corrective systems. The same data architecture may serve climate adaptation, insurance, housing, identity and security; separability prevents convenience at the infrastructure layer from becoming total social classification. [[articles/Climate Meritocracy|How Reparative Justice Became Meritocracy]] shows why the safeguard is necessary. Publicly produced environmental observations can acquire consequences far beyond their scientific purpose once private models connect exposure to credit, investment and insurability. ## Person of Interest Context The relevant/irrelevant division in [[wiki/Person of Interest|Person of Interest]] is a dramatic example of a classification becoming a constitution. [[wiki/The Machine|The Machine]] and [[wiki/Samaritan|Samaritan]] show how a score or category can allocate attention, protection and coercion before affected people know the classification exists. ## Requirements - purpose limitation and separate data stores where appropriate; - explicit prohibition on cross-domain reuse without new authority; - independent review and [[wiki/Model Contestability|Model Contestability]]; - explanation of which domain produced a consequential determination; - correction that propagates to every authorized copy; and - rights and minimum services that remain non-scoreable. ## Key Insight **Advanced governance must distinguish risk, need, contribution, reliability, entitlement and dignity rather than collapsing them into one administrative identity.** ## See Also [[wiki/Risk Scoring Systems|Risk Scoring Systems]], [[wiki/Algorithmic Constitutionalism|Algorithmic Constitutionalism]], [[wiki/Appeals Mechanisms|Appeals Mechanisms]], [[wiki/Dual Constitution of AI|Dual Constitution of AI]], [[wiki/Rights|Rights]]