GOVERNATLAS / PRODUCT DOSSIER

InterpretML

InterpretML documents tools for fitting interpretable models and explaining black-box machine-learning systems.

Explore the product ↗Add to comparison ↗

01 / CONTEXT & SCOPE

Understand the proposition.
Define the boundary.

Where it fits

Distinguish an intrinsically interpretable model from an explanation applied to a different model.

The decision to make

Separate an interpretable model fitted with the package from a post-hoc explanation of a different model. The explanation contract, preprocessing and validation responsibility differ. A global model view also cannot explain every individual workflow decision without the served input, output and downstream rule.

What this record does not establish

Documented interpretability methods and example benchmark tables do not establish causal effects, fair decisions or equivalent accuracy in a buyer environment. Any privacy variant and its assumptions require separate review.

1 current source · 4 published assertions
Source coverage is a research boundary, not a quality score.

02 / YOUR EVALUATION

Turn interest into a useful trial.

Open the evaluation playbook ↗

A test worth running

Fit or inspect a small synthetic additive routing model in an authorized environment, then retain a local explanation and a global summary. Change one preprocessing field in a separate diagnostic and show which artifact no longer matches the served configuration. Review a surprising case without relying on the aggregate view.

GovernAtlas editorial proposal. We have not performed this test.

Evidence to bring back

A model-versus-explainer/preprocessing/output-scale manifest.
A local/global comparison with a deliberately mismatched configuration identified.

THE OPERATING DETAIL

Know which model the explanation belongs to.

GovernAtlas analysis: practical review considerations, not findings from a product test.

Choose the object being reviewed

Document whether the trial fits a glassbox model or explains an existing black-box system. Preserve the served estimator, input transformations and output scale. A clear explanation of a surrogate or alternate model does not automatically describe the predictions made by the production model.

Connect local and global views

Use one fictional misclassified case and a documented reference dataset. Compare its local explanation with the model’s global behavior. Inspect an interaction or unusual feature range that a summary could conceal. Ask a second reviewer to explain the difference without treating importance as causation.

Retest an expert edit

If an expert proposes a change to an interpretable model, retain the original model and rationale. Evaluate the amended model on held-out cases and an unaffected control. Legibility makes a change easier to inspect; it does not establish that the edit improves outcomes or preserves all earlier behavior.

03 / DOCUMENTED EVIDENCE

Read the claim. Inspect the basis.

Each statement retains its classification and preserved source. A vendor assertion is not an independent test result.

first party asserted

The README demonstrates global model explanations, local prediction explanations and comparison of explanation objects.

Source, dates & preserved provenance

InterpretML — primary scope and current identity ↗

Authority: first party
Observed 2026-10-16 · Reviewed 2026-10-16 · current until 2026-11-15

Artifact SHA-256: 1b5ee328c74ae38f117095a92e2d05b2122d2da3c19a73c928dbe849f89d3b7d

The hash identifies the preserved capture. The live source may have changed.

first party asserted

Its listed black-box explanation methods include SHAP Kernel Explainer, LIME, sensitivity analysis and partial dependence.

Source, dates & preserved provenance

InterpretML — primary scope and current identity ↗

Authority: first party
Observed 2026-10-16 · Reviewed 2026-10-16 · current until 2026-11-15

Artifact SHA-256: 1b5ee328c74ae38f117095a92e2d05b2122d2da3c19a73c928dbe849f89d3b7d

The hash identifies the preserved capture. The live source may have changed.

first party asserted

InterpretML documents training glassbox models, including Explainable Boosting Machines, decision trees and linear or logistic regression.

Source, dates & preserved provenance

InterpretML — primary scope and current identity ↗

Authority: first party
Observed 2026-10-16 · Reviewed 2026-10-16 · current until 2026-11-15

Artifact SHA-256: 1b5ee328c74ae38f117095a92e2d05b2122d2da3c19a73c928dbe849f89d3b7d

The hash identifies the preserved capture. The live source may have changed.

first party asserted

InterpretML documents tools for fitting interpretable models and explaining black-box machine-learning systems. Scope: primary documentation only; no independent model test, fairness determination, causal finding or compliance approval.

Source, dates & preserved provenance

InterpretML — primary scope and current identity ↗

Authority: first party
Observed 2026-10-16 · Reviewed 2026-10-16 · current until 2026-11-15

Artifact SHA-256: 1b5ee328c74ae38f117095a92e2d05b2122d2da3c19a73c928dbe849f89d3b7d

The hash identifies the preserved capture. The live source may have changed.

Source register

InterpretML — primary scope and current identity ↗first party · captured 2026-10-16

Brand asset provenance

Organization identity asset from its official website; inclusion does not imply endorsement.

Official website ↗ · Original asset ↗

04 / BEFORE YOU COMMIT

Better questions.
A clearer decision.

Resolve these with the provider and your own evaluation team.

Can you provide a model-versus-explainer/preprocessing/output-scale manifest for the selected configuration?

Can you provide a local/global comparison with a deliberately mismatched configuration identified for the selected configuration?

Which exact version, entitlements, data paths and responsible roles apply?

Regulatory applicability requires a separate, current determination. Inclusion is not certification.

COMPLIANCE / CLAIMS, SCOPE & GAPS

Start with the requirement.
Inspect the support.

Prepare your vendor review ↗

These labels describe this dossier’s current evidence coverage. A published mention may be a framework mapping, a qualified assertion or a limitation; it is not a compliance finding.