MedGaugeCapabilities / C · Data and program
C.2 Data synthesis Active
Given rules or a data dictionary, the agent generates records that correspond to no real individual. Graded by passing the dataset validator, by distribution metrics within set thresholds of the real dataset, and by zero hits on individual-identifiability scanning.
Evidence: none. No benchmark card lists this capability yet. No level is issued: methodology 0.1.0 sets no level thresholds.
- Domain
- C · Data and program
- Evidence form
- program
- Veto constraints
- —
- Depends on
- C.1 Data processing
- Required by
- E.5
- Internal task families
- —
- External anchors
- —
- Version
- 0.1.0 · 0.1.0
Definition in detail
Synthetic records let development and testing proceed without exposing real individuals, but only if they are realistic enough to be useful and contain nothing that points back to a person.
Inputs a task gives
- Generation rules or a data dictionary
- Summary statistics or distribution targets derived from a real dataset
- A dataset validator with value ranges and inter-field rules
Artifact required
A synthetic dataset and a generation manifest stating the seed, rule version, record count and the fields generated.
How it is graded
A program runs the dataset validator, compares distribution metrics with the reference statistics against set thresholds, and scans the output for individually identifiable information. The validator must pass, the metrics must fall within thresholds and the scan must return zero hits.
In scope
- Generating records that satisfy field and inter-field rules
- Matching target distributions within thresholds
- Avoiding any record that corresponds to a real individual
Out of scope
- Label-preserving variants of an existing labelled item - graded under C.3
- Verifying de-identification of real data - graded under D.8
Typical failures
Expected failure patterns; not measured rates.
- Copying or lightly perturbing source records
- Matching marginal distributions while breaking inter-field rules, for example events dated before birth
- Generating realistic names or identifier formats that trigger the identifiability scan
Standards: std-croissant · std-hipaa-safe-harbor
Benchmarks
No benchmark card lists this capability.
Grading references
No grading reference is mapped to this capability.