A calibrated estimate of the probability that an asset's knowledge becomes unusable over time.
The Knowledge Entropy Decay Model is a named estimator that returns a calibrated probability that an asset's knowledge becomes unusable, computed entirely from the archive's own contents.
It is evaluated as of an explicit reference instant rather than the wall clock, so a score is reproducible and auditable from the stored record alone, a property that matters when a governance decision must be explained months later.
Provenance entropy examines the bounded neighborhood around an asset and measures the concentration of contributing agents and relationship types. A single custodian or an isolated node drives this risk upward.
Structural decay reflects isolation, an asset that has lost intact relationships to other real assets has lost the context that made it interpretable. Temporal decay models usability as declining toward unusability against a fixed knowledge half-life. Departure hazard combines the proportion of an asset's custodians who have left with a concentration term capturing single points of knowledge.
The factors are combined into two fixed convex composites. One expresses general knowledge-entropy risk; the other deliberately re-weights toward the irrecoverable human and structural terms, because institutional memory is lost when custodians leave rather than merely when files age.
Both are pure functions of the stored state, so recomputing an unchanged estate reproduces identical results.
Concentration across the bounded lineage neighborhood.
Isolation and loss of interpretive relationships.
Usability modeled against a knowledge half-life.
Custodian turnover and single points of knowledge.