AI0-001 AI Implementation and Operations Practice Question

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  Publish the full training dataset and model weights so applicants can inspect them directly.

  Why it's wrong here

  Releasing training data would expose other customers' personal information and violate privacy obligations, and raw weights are not meaningful to an applicant. Transparency duties call for understandable explanations of the decision, not disclosure of proprietary or personal data. This practice would create legal exposure rather than satisfy the contestability requirement.

• ✓

  Generate per-applicant reason codes that identify the principal factors driving each decision, using techniques such as SHAP or LIME.

  Why this is correct

  Reason codes translate a model score into the specific factors, such as debt-to-income ratio or recent delinquencies, that moved the decision. This gives applicants the meaningful information regulators expect and gives reviewers a concrete basis for evaluating a contest. SHAP and LIME produce these local attributions from the deployed model without altering its predictions.

• ✗

  Disable logging of decision inputs to minimize the personal data retained about applicants.

  Why it's wrong here

  Without logged inputs and outputs there is no way to reconstruct a decision, verify the reason codes, or support a review request. Minimizing retention is a sound privacy principle, but eliminating decision logs defeats both transparency and contestability. Retention should be scoped and time-limited, not removed entirely for decisions that must be explainable.

• ✗

  Reduce the model to a single decision tree so every applicant can be shown the same global tree structure.

  Why it's wrong here

  Forcing a simpler architecture may lower accuracy and still does not deliver per-applicant explanations, since a global tree does not show which path a specific applicant followed or why. Interpretability requirements are met through local attributions and review processes, not by degrading model performance for all users in the hope that a shared structure suffices.

• ✓

  Provide a documented human review pathway where applicants can request reconsideration and a qualified reviewer can override the automated decision.

  Why this is correct

  Contestability requires an actionable route for an applicant to challenge an outcome and obtain human judgment. A documented review process with trained reviewers who can examine the reason codes and override the model ensures the automated decision is not final and unreviewable. This is the operational mechanism that makes the right to contest meaningful.