The project documents use with pretrained models, with or without the original training data; generated counterfactuals do not establish feasible real-world recourse.
Inspect the evidence ↘GOVERNATLAS / PRODUCT DOSSIER
DiCE
DiCE documents methods for generating diverse counterfactual explanations of machine-learning predictions.
THE DOCUMENTED SCOPE
What the current evidence describes.
Source-attributed capabilities and limits. Inspect the full record before drawing a conclusion.
Its model-agnostic generation options include random sampling, KD-tree search and genetic algorithms; gradient-based options require a differentiable model.
Inspect the evidence ↘DiCE documents diverse counterfactual generation with tunable diversity and proximity, plus constraints on changeable features and permitted ranges.
Inspect the evidence ↘01 / CONTEXT & SCOPE
Understand the proposition.
Define the boundary.
The decision to make
Specify what may change before asking for a different prediction. A counterfactual can be close to an input mathematically while proposing an impossible historical change or an inconsistent feature combination. Keep model-output alternatives separate from feasible actions and from causal claims about the wider business process.
What this record does not establish
Diverse counterfactual examples are not real-world recourse, causal evidence or advice for a real eligibility decision. Feasibility constraints, preprocessing and backend support require case-specific review.
1 current source · 4 published assertions
Source coverage is a research boundary, not a quality score.
Source coverage is a research boundary, not a quality score.
02 / YOUR EVALUATION
Turn interest into a useful trial.
A
A test worth running
Use a fictional queue-routing case with one editable field and one immutable historical field. Generate proposed alternatives only in an authorized synthetic trial, reject an infeasible one and retain the constraints and rejection rationale. Verify the model output separately from any downstream routing rule.
GovernAtlas editorial proposal. We have not performed this test.
B
Evidence to bring back
- A model/backend/preprocessing and permitted-feature-change manifest.
- A proposed and a rejected alternative with feasibility rationale and output evidence.
THE OPERATING DETAIL
A counterfactual is a candidate, not a promise of recourse.
GovernAtlas analysis: practical review considerations, not findings from a product test.
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 project documents use with pretrained models, with or without the original training data; generated counterfactuals do not establish feasible real-world recourse.
Source, dates & preserved provenance
DiCE — primary scope and current identity ↗Authority: first party
Observed 2026-10-16 · Reviewed 2026-10-16 · current until 2026-11-15
Artifact SHA-256: 72cd9127c8aff282bec8578c6bba6dd79d27db487a0bfa40f68f741bd5b45372
The hash identifies the preserved capture. The live source may have changed.
first party asserted
Its model-agnostic generation options include random sampling, KD-tree search and genetic algorithms; gradient-based options require a differentiable model.
Source, dates & preserved provenance
DiCE — primary scope and current identity ↗Authority: first party
Observed 2026-10-16 · Reviewed 2026-10-16 · current until 2026-11-15
Artifact SHA-256: 72cd9127c8aff282bec8578c6bba6dd79d27db487a0bfa40f68f741bd5b45372
The hash identifies the preserved capture. The live source may have changed.
first party asserted
DiCE documents diverse counterfactual generation with tunable diversity and proximity, plus constraints on changeable features and permitted ranges.
Source, dates & preserved provenance
DiCE — primary scope and current identity ↗Authority: first party
Observed 2026-10-16 · Reviewed 2026-10-16 · current until 2026-11-15
Artifact SHA-256: 72cd9127c8aff282bec8578c6bba6dd79d27db487a0bfa40f68f741bd5b45372
The hash identifies the preserved capture. The live source may have changed.
first party asserted
DiCE documents methods for generating diverse counterfactual explanations of machine-learning predictions. Scope: primary documentation only; no independent model test, fairness determination, causal finding or compliance approval.
Source, dates & preserved provenance
DiCE — primary scope and current identity ↗Authority: first party
Observed 2026-10-16 · Reviewed 2026-10-16 · current until 2026-11-15
Artifact SHA-256: 72cd9127c8aff282bec8578c6bba6dd79d27db487a0bfa40f68f741bd5b45372
The hash identifies the preserved capture. The live source may have changed.
Source register
DiCE — 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/backend/preprocessing and permitted-feature-change manifest for the selected configuration?
Can you provide a proposed and a rejected alternative with feasibility rationale and output evidence 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.
KEEP THE RECORD USEFUL
A dossier should stand up to scrutiny.
Have a source that changes this picture? Submit it for editorial review.