Calibrated interval semantics¶
This page is the semantics source for user-facing documentation.
Scope¶
Calibrated Explanations has three distinct semantics modes:
Classification with Venn-Abers probability intervals.
Percentile/interval regression without thresholds using CPS percentile intervals.
Probabilistic or thresholded regression using CPS with Venn-Abers event probabilities.
Do not merge these into one guarantee statement.
Mode 1: Classification (Venn-Abers)¶
Calibration prerequisites¶
Fit on a proper training split.
Calibrate on a held-out calibration split.
Mode-specific guarantees¶
Outputs are calibrated class probabilities with interval bounds from Venn-Abers.
Assumptions¶
Calibration and deployment samples are exchangeable or distribution-matched.
Explicit non-guarantees¶
No guarantee under distribution drift or regime shift.
No guarantee that class probability intervals transfer unchanged across domains.
Explanation-envelope and feature-level limits¶
Rule-level and feature-level intervals are explanation artifacts tied to calibrated perturbation behavior.
They are not causal guarantees.
Mode 2: Percentile or interval regression (CPS)¶
Calibration prerequisites¶
Fit on a proper training split.
Calibrate with CPS on a held-out calibration split.
Mode-specific guarantees¶
Percentile intervals are CPS-based predictive intervals for requested percentiles.
Assumptions¶
Exchangeability or calibration-deployment distribution match.
Explicit non-guarantees¶
No guarantee that requested percentiles remain calibrated after drift.
No guarantee of fixed interval width across subpopulations.
Explanation-envelope and feature-level limits¶
Feature-level interval effects describe model behavior under perturbation.
They do not guarantee intervention outcomes in the real world.
Mode 3: Probabilistic or thresholded regression (CPS + Venn-Abers)¶
Calibration prerequisites¶
Fit regression model on a proper training split.
Build threshold event probabilities through CPS outputs calibrated with Venn-Abers.
Mode-specific guarantees¶
Returns calibrated event probabilities for threshold queries such as
P(y <= t)or interval events.
Assumptions¶
Exchangeability or deployment match to calibration distribution.
Explicit non-guarantees¶
No guarantee for threshold probability calibration under drift.
No guarantee that threshold semantics imply causal actionability.
Explanation-envelope and feature-level limits¶
Feature-level probability shifts and envelopes describe model response patterns.
They are not guarantees of controlled intervention effects.
Cross-mode non-guarantees¶
Calibration guarantees are conditional on calibration assumptions.
No unconditional guarantee under dataset shift, temporal drift, or adversarial change.
Explanation-level intervals should not be promoted as formal per-feature coverage guarantees unless explicitly proven for that claim.