Schema
The structural constraints — types, relationships, integrity rules — enforced at the data layer before any application logic runs.
Data quality is assessed through application-layer validation: error rates, failed transactions, and audit logs that capture what went wrong after the fact — measures that document the downstream effects of structural problems without touching the structural layer where those problems originate.
The schema itself as a governance artifact: whether data types encode structural assumptions about identity or classification, whether integrity rules prevent historically discriminatory data patterns from persisting, and whether schema changes are subject to any review process before they propagate through downstream systems.
When the schema becomes a governance artifact rather than a technical implementation detail, algorithmic accountability reaches the layer that actually shapes what can be stored, compared, and inferred — not only the layer that reports what was computed.
Civil rights advocates, data privacy regulators, and individuals subject to automated classification who gain standing to challenge the structural encoding of categories before those categories propagate through downstream models and decisions.
Database vendors, data brokers, and organizations that profit from the unrestricted accumulation and combination of personal data resist schema-level governance because it would require them to justify the structural decisions that currently operate below the threshold of regulatory visibility.
A data governance framework that regulates what organizations do with data without auditing the schema layer that determines what can be stored and inferred is regulating the output of a press without examining the mold — the structural constraints that shape every impression downstream.
Under CCPA and GDPR, you have the right to a machine-readable copy of your personal data. When you receive it, look at the field names and categories — unexpected fields are legitimate subjects of a follow-up inquiry.
Data minimization rules operate at the schema level even when described in policy terms. Submit a comment to your state attorney general or to the FTC's rulemaking record asking that schema design be explicitly covered under data minimization requirements.
HR systems, student information systems, and workplace monitoring platforms all have schemas. Under relevant privacy laws, you can request the categories of personal information maintained about you — and the response tells you something about the structural decisions that govern your data.
Organizations like the Open Data Institute (theodi.org) and Schema.org work on public schema standards that make data structurally legible across systems. Supporting their work directly advances the norm that structural data decisions should be transparent and standardized.