Inputs
- Evidence quality signals
- Decision context
- Post-solve analysis inputs
Control and decision
Confidence Estimator
Uncalibrated confidence, stated as uncalibrated.
In-process evidence-sufficiency scoring and an explicitly uncalibrated confidence object attached to decision metadata.
Responsibility
What Confidence Estimator is responsible for
The Confidence Estimator module in the Control Plane records how sufficient the available evidence is for a decision and attaches a confidence object to decision metadata. In the reviewed implementation that object is explicitly uncalibrated: it reports calibrated:false, probability 0 and method "none" rather than inventing a risk score.
Most systems encode uncertainty in an implicit threshold: a model returns a score, a rule cuts it, and confidence becomes invisible. NiGeNje's design intent is the opposite — uncertainty should be an input a decision can reason about and an auditor can inspect.
What exists today is more modest and more honest than a risk engine: an evidence-sufficiency score and a confidence object that declares itself uncalibrated. No calibrated probability, threshold ownership or control effect is claimed.
Lifecycle position
Where it sits in the governance lifecycle
Stage numbering is canonical across the site. Open the interactive lifecycle to see every component with responsibilities at each stage.
Decision
Coordinate policy, constraints and uncertainty.
Verified capabilities
What is verified today
Capabilities are drawn from the product register. Each one is bounded by the limitations stated further down this page.
- Evidence-sufficiency scoring for a decision context (in-process)
- Explicit uncalibrated confidence metadata (calibrated:false, probability 0, method none)
- Post-solve chance and CVaR analysis in the IDM path (bounded, synthetic)
Outputs
- Evidence-sufficiency finding
- Explicitly uncalibrated confidence metadata
- Explicit unknown when confidence cannot be established
Evidence and maturity
What the evidence supports
Maturity describes verified engineering evidence in a bounded environment. It is not a production-readiness statement.
Maturity note
An in-process evidence-sufficiency and confidence-metadata module is present in the Control Plane. It does not own thresholds or influence control decisions, and it explicitly reports an uncalibrated state.
Not production-qualified
This label describes verified engineering evidence only. No production deployment, certification, benchmark or customer outcome is claimed.
Evidence recorded
- Control Plane engines/model source: evidence-sufficiency and confidence metadata with calibrated:false.
- Confidence/AIOps scheduling report documenting the uncalibrated state.
- Known-limitations register entry 7 records synthetic CLF/CBF and post-solve chance/CVaR handling.
Dependencies
- Control Plane
- IDM Solver
- evidence records
Dependencies are stated as the product register records them. A dependency is not automatically production-qualified; see its own product or component page for its maturity and limitations.
Limitations
What this product does not do (yet).
Limitations are stated before the evaluation questions so they cannot be missed, downplayed or treated as future-work footnotes.
Limitations — read before evaluating
- No calibrated estimator, probability model or risk score is implemented in the reviewed code.
- The module is in-process Control Plane metadata; it owns no threshold and has no control effect.
- No calibration, accuracy or benchmark claim is made.
- Where uncertainty cannot be estimated, the honest state is unknown — not a default confidence value.
No capability on this page should be read as production qualification. If a limitation is material to your use case, it belongs in the evaluation scope — not after it.
Relationship to the platform
How it fits
The estimator, where present, informs but does not authorise. It cannot issue a permit, and a high confidence value does not substitute for missing evidence.
Evaluation questions
- What evidence does the risk signal depend on?
- How are uncalibrated estimates prevented from becoming implicit authorisation?
- Is uncertainty visible to the reviewer or collapsed into a boolean?
Next step
Evaluate Confidence Estimator against a real control problem.
Bring a system boundary, an action class and a failure mode. The first step is to agree what should be governed and what evidence would count.