Uncertainty Lab
Make aleatoric and epistemic uncertainty, distribution shift, evidence quality, and abstention visible before any recommendation.
WHAT THIS LAB ESTABLISHES
High confidence can coexist with high ignorance.
Uncertainty is displayed before the output, and abstention remains a successful system state.
INSPECTABLE REPRESENTATION
Visual, table, and text views
| Field | State | Qualification |
|---|---|---|
| Scenario | Synthetic anomaly classification | Fictional or abstract |
| Output | Derived conditions | No guilt, blame, legal liability, targeting, or universal score |
| Persistence | None for public simulations | Human-judgment interaction remains browser-local |
Text description
Make aleatoric and epistemic uncertainty, distribution shift, evidence quality, and abstention visible before any recommendation.
The demonstration uses synthetic anomaly classification. It exposes qualifications and preserves unknown states. The output is not a legal conclusion, operational recommendation, force authorization, or certification.
DERIVED RESULT
Review conditions and qualifications
A high model output can coexist with high epistemic uncertainty. Numerical confidence is not proof, authority, or evidential sufficiency.