Glossary · Information modeling
Semantic drift
In information architecture, semantic drift is a change in the meaning or interpretation of a term, signal, state, field or value while its label or structure appears stable.
- Information architecture
- Ontology
In one sentence
Semantic drift is when a term, field or signal quietly changes meaning while its name stays the same; schema checks cannot see it.
Example
A PLC tag named MotorOK originally meant “drive ready”; after a software update it means “no active fault,” while the HMI and the operating manual still describe the old meaning.
How it applies
- Integrated systems: Drift typically appears when one side changes units, reference points, state definitions or thresholds without renaming the field. Schema validation cannot detect it, because the structure is unchanged.
- Functional safety: Consumers that act on a drifted value do the wrong thing with full confidence. Treat changes of meaning as interface changes, analyze their change propagation, and prevent drift through semantic control.
- Technical documentation: Documentation drifts too: instructions, glossaries and warning messages keep a term whose meaning in the product has changed. See information drift.
- AI and retrieval: Language models and search indexes learn terms from historical content. If a meaning has shifted, old and new usage get mixed; stable definitions with IRIs and review dates reduce this.
Semantic drift vs. a documented change
A change of meaning is not drift if it is announced, versioned and reflected in every consumer and document. It becomes drift when the label stays the same and nobody is told.