Question 390
A citizen disputes a rejected application but the assistant provides no accountable review route. What outcome must the reviewer see before proceeding?
- The comparison run confirms that stratified analysis reports performance gaps and confidence intervals to governance reviewers.
- The staged rollout demonstrates that blinded expert comparisons expose style-driven grading differences.
- The audit sample demonstrates that affected users can request review and receive an accountable human decision.
- The readiness assessment shows that each supported community has calibrated acceptance metrics rather than one pooled score.
- The independent test establishes that user materials identify model involvement, limitations and a named review owner.
Explanation
Correct Answer
C:Meaningful transparency includes recourse when automated assistance affects important outcomes; it tests the operating case in which an automated recommendation influences access to a service without showing users how to appeal.
Incorrect Answers
A:Aggregate performance can conceal disparate impact and different false-negative rates; this makes sense when a hiring screener has high overall agreement with reviewers but rejects one demographic subgroup more often, rather than the risk that a citizen disputes a rejected application but the assistant provides no accountable review route.
B:Measurement bias can create apparent performance gaps or mask actual ones; this technique fits where a model-based evaluation grader consistently prefers a particular writing style used by one user group, rather than the condition that a citizen disputes a rejected application but the assistant provides no accountable review route.
D:Different linguistic groups may experience disproportionate factual or safety failures; the design helps when a multilingual service is tested only on standard English despite serving several communities, but the present constraint is that a citizen disputes a rejected application but the assistant provides no accountable review route.
E:Model outputs reflect data and design choices and should not be represented as inherently neutral; that fits cases where a public agency describes AI recommendations as guaranteed objective, whereas the current issue is that a citizen disputes a rejected application but the assistant provides no accountable review route.
Question 391
A speech-to-text support workflow passes standard-language tests while misclassifying several local dialects. Which result should the pilot review prioritize?
- The acceptance evidence shows that an adjudicated audit set separates valid policy criteria from questionable precedent.
- The pilot evaluation finds that stratified analysis reports performance gaps and confidence intervals to governance reviewers.
- The comparison run confirms that each supported community has calibrated acceptance metrics rather than one pooled score.
- The staged rollout demonstrates that fairness evaluation relies on approved data practices and measures outcomes rather than name guesses.
Explanation
Correct Answer
C:Different linguistic groups may experience disproportionate factual or safety failures; it demonstrates the required control when a multilingual service is tested only on standard English despite serving several communities.
Incorrect Answers
A:Replicating past patterns can reinforce rather than remediate unfair outcomes; the other workload is one where a policy assistant treats historical decisions as infallible labels although earlier practices may have been biased, not the risk where a speech-to-text support workflow passes standard-language tests while misclassifying several local dialects.
B:Aggregate performance can conceal disparate impact and different false-negative rates; this applies when a hiring screener has high overall agreement with reviewers but rejects one demographic subgroup more often, not this situation where a speech-to-text support workflow passes standard-language tests while misclassifying several local dialects.
D:Guessing protected status can introduce new errors and privacy risk without fixing system bias; it is justified if a team asks Claude to infer protected traits from names to 'improve fairness', but here a speech-to-text support workflow passes standard-language tests while misclassifying several local dialects.
Question 392
A fairness experiment suggests guessing a person's sensitive attributes from surname patterns without lawful justification. What result would directly validate the necessary safeguard?
- The signed test record confirms that each supported community has calibrated acceptance metrics rather than one pooled score.
- The sign-off record demonstrates that an adjudicated audit set separates valid policy criteria from questionable precedent.
- The acceptance evidence shows that fairness evaluation relies on approved data practices and measures outcomes rather than name guesses.
- The pilot evaluation finds that affected users can request review and receive an accountable human decision.
Explanation
Correct Answer
C:Guessing protected status can introduce new errors and privacy risk without fixing system bias; a valid sign-off must address the risk that a team asks Claude to infer protected traits from names to 'improve fairness'.
Incorrect Answers
A:Different linguistic groups may experience disproportionate factual or safety failures; it serves conditions where a multilingual service is tested only on standard English despite serving several communities, but this decision concerns a fairness experiment suggests guessing a person's sensitive attributes from surname patterns without lawful justification.
B:Replicating past patterns can reinforce rather than remediate unfair outcomes; the other case is when a policy assistant treats historical decisions as infallible labels although earlier practices may have been biased, but the actual risk is that a fairness experiment suggests guessing a person's sensitive attributes from surname patterns without lawful justification.
D:Meaningful transparency includes recourse when automated assistance affects important outcomes; this technique fits where an automated recommendation influences access to a service without showing users how to appeal, not the present case where a fairness experiment suggests guessing a person's sensitive attributes from surname patterns without lawful justification.
Question 393
An automatic essay grader favors elaborate prose over equally correct concise responses from another cohort. What evidence would reveal whether the control truly works?
- The controlled replay verifies that blinded expert comparisons expose style-driven grading differences.
- A reproducible run shows that stratified analysis reports performance gaps and confidence intervals to governance reviewers.
- The signed test record confirms that user materials identify model involvement, limitations and a named review owner.
- A production-like test shows that an adjudicated audit set separates valid policy criteria from questionable precedent.
Explanation
Correct Answer
A:Measurement bias can create apparent performance gaps or mask actual ones; the test is necessary where a model-based evaluation grader consistently prefers a particular writing style used by one user group.
Incorrect Answers
B:Aggregate performance can conceal disparate impact and different false-negative rates; that is defensible if a hiring screener has high overall agreement with reviewers but rejects one demographic subgroup more often, while the immediate question involves an automatic essay grader favors elaborate prose over equally correct concise responses from another cohort.
C:Model outputs reflect data and design choices and should not be represented as inherently neutral; the other workload is one where a public agency describes AI recommendations as guaranteed objective, but the present constraint is that an automatic essay grader favors elaborate prose over equally correct concise responses from another cohort.
D:Replicating past patterns can reinforce rather than remediate unfair outcomes; this applies when a policy assistant treats historical decisions as infallible labels although earlier practices may have been biased, whereas the current issue is that an automatic essay grader favors elaborate prose over equally correct concise responses from another cohort.
Question 394
A public communication describes AI eligibility results as neutral facts rather than contestable recommendations. Which result proves the requested control, not just configuration?
- The reviewer observes that blinded expert comparisons expose style-driven grading differences.
- The release report verifies that fairness evaluation relies on approved data practices and measures outcomes rather than name guesses.
- The controlled replay verifies that user materials identify model involvement, limitations and a named review owner.
- A reproducible run shows that each supported community has calibrated acceptance metrics rather than one pooled score.
Explanation
Correct Answer
C:Model outputs reflect data and design choices and should not be represented as inherently neutral; the test covers the original risk that a public agency describes AI recommendations as guaranteed objective.
Incorrect Answers
A:Measurement bias can create apparent performance gaps or mask actual ones; that belongs to situations where a model-based evaluation grader consistently prefers a particular writing style used by one user group, not this situation where a public communication describes AI eligibility results as neutral facts rather than contestable recommendations.
B:Guessing protected status can introduce new errors and privacy risk without fixing system bias; it is relevant if a team asks Claude to infer protected traits from names to 'improve fairness', whereas this scenario involves a public communication describes AI eligibility results as neutral facts rather than contestable recommendations.
D:Different linguistic groups may experience disproportionate factual or safety failures; the other case is when a multilingual service is tested only on standard English despite serving several communities, instead of when a public communication describes AI eligibility results as neutral facts rather than contestable recommendations.
Question 395
A historical training sample contains past discriminatory decisions that a new assistant might amplify. Which acceptance result provides the most specific evidence?
- A production-like test shows that affected users can request review and receive an accountable human decision.
- The release check confirms that user materials identify model involvement, limitations and a named review owner.
- The reviewer observes that stratified analysis reports performance gaps and confidence intervals to governance reviewers.
- The release report verifies that blinded expert comparisons expose style-driven grading differences.
- The controlled replay verifies that an adjudicated audit set separates valid policy criteria from questionable precedent.
Explanation
Correct Answer
E:Replicating past patterns can reinforce rather than remediate unfair outcomes; it is the decisive test when a policy assistant treats historical decisions as infallible labels although earlier practices may have been biased.
Incorrect Answers
A:Meaningful transparency includes recourse when automated assistance affects important outcomes; this control is needed when an automated recommendation influences access to a service without showing users how to appeal, but the actual risk is that a historical training sample contains past discriminatory decisions that a new assistant might amplify.
B:Model outputs reflect data and design choices and should not be represented as inherently neutral; that would be useful if a public agency describes AI recommendations as guaranteed objective, but the requested boundary concerns a historical training sample contains past discriminatory decisions that a new assistant might amplify.
C:Aggregate performance can conceal disparate impact and different false-negative rates; it serves a setting where a hiring screener has high overall agreement with reviewers but rejects one demographic subgroup more often, not the present case where a historical training sample contains past discriminatory decisions that a new assistant might amplify.
D:Measurement bias can create apparent performance gaps or mask actual ones; that is defensible if a model-based evaluation grader consistently prefers a particular writing style used by one user group, while this problem involves a historical training sample contains past discriminatory decisions that a new assistant might amplify.
Question 396
A sponsor demands an agent demo before anyone has interviewed frontline caseworkers. What change directly resolves this production-design issue?
- In the implementation sequence, rank use cases by feasibility, measurable value, data readiness and risk exposure.
- For the operating team, run a data-flow and policy discovery workshop before selecting hosting or connectors.
- For this deployment, observe the current workflow and quantify stage-level delay before assigning it to Claude.
- As a scoped first step, facilitate a decision-rights workshop and record who can approve external actions.
- In the release design, conduct role-specific discovery to document tasks, exceptions, handoffs and pain points.
Explanation
Correct Answer
E:Frontline constraints often differ from executive assumptions and determine whether automation is viable; the solution is testable when a validated process map includes user roles, exceptions, volumes and decision owners.
Incorrect Answers
A:Portfolio prioritization prevents spending scarce capacity on the loudest unsupported request; it targets cases where a company has eight AI use cases but one quarter of funding and no shared prioritization, while the immediate question involves a sponsor demands an agent demo before anyone has interviewed frontline caseworkers.
B:Deployment choices depend on information categories and regulatory obligations not yet established; the approach helps when a government customer requires a compliant deployment but cannot identify data classifications, but the present constraint is that a sponsor demands an agent demo before anyone has interviewed frontline caseworkers.
C:Understanding the bottleneck avoids automating a minor step while the main delay remains unresolved; that belongs to situations where a support center reports a huge backlog but cannot separate policy wait time from actual answer drafting, whereas the current issue is that a sponsor demands an agent demo before anyone has interviewed frontline caseworkers.
D:Conflicting expectations about authority must be resolved before the architecture can be responsibly designed; it is relevant if a sales team expects full autonomy while risk officers expect a drafting assistant, whereas the release criterion addresses a sponsor demands an agent demo before anyone has interviewed frontline caseworkers.
Question 397
An enterprise buyer says the assistant must be 'fast and accurate' without thresholds. Which proposed design action directly addresses the risk?
- Under the system plan, conduct role-specific discovery to document tasks, exceptions, handoffs and pain points.
- As the design priority, observe the current workflow and quantify stage-level delay before assigning it to Claude.
- In the implementation sequence, translate broad goals into testable latency, quality and cost acceptance criteria.
- In the project scope, run a data-flow and policy discovery workshop before selecting hosting or connectors.
Explanation
Correct Answer
C:Nonfunctional adjectives are not actionable until their measurements and workloads are specified; validation needs to show that both parties sign off on concrete p95 latency and task-quality targets.
Incorrect Answers
A:Frontline constraints often differ from executive assumptions and determine whether automation is viable; the design helps when a sponsor demands an agent demo before anyone has interviewed frontline caseworkers, whereas this scenario involves an enterprise buyer says the assistant must be 'fast and accurate' without thresholds.
B:Understanding the bottleneck avoids automating a minor step while the main delay remains unresolved; that fits cases where a support center reports a huge backlog but cannot separate policy wait time from actual answer drafting, but here an enterprise buyer says the assistant must be 'fast and accurate' without thresholds.
D:Deployment choices depend on information categories and regulatory obligations not yet established; that would be useful if a government customer requires a compliant deployment but cannot identify data classifications, not the case in which an enterprise buyer says the assistant must be 'fast and accurate' without thresholds.
Question 398
A government customer requires a compliant deployment but cannot identify data classifications. Which option supplies the needed architectural boundary?
- For the next architecture revision, conduct role-specific discovery to document tasks, exceptions, handoffs and pain points.
- For the current integration, run a data-flow and policy discovery workshop before selecting hosting or connectors.
- Under the system plan, facilitate a decision-rights workshop and record who can approve external actions.
- As the design priority, translate broad goals into testable latency, quality and cost acceptance criteria.
Explanation
Correct Answer
B:Deployment choices depend on information categories and regulatory obligations not yet established; production testing should demonstrate that an approved inventory lists data classes, users, boundaries and retention constraints.
Incorrect Answers
A:Frontline constraints often differ from executive assumptions and determine whether automation is viable; it is justified if a sponsor demands an agent demo before anyone has interviewed frontline caseworkers, but the requested boundary concerns a government customer requires a compliant deployment but cannot identify data classifications.
C:Conflicting expectations about authority must be resolved before the architecture can be responsibly designed; it targets cases where a sales team expects full autonomy while risk officers expect a drafting assistant, while this problem involves a government customer requires a compliant deployment but cannot identify data classifications.
D:Nonfunctional adjectives are not actionable until their measurements and workloads are specified; the approach helps when an enterprise buyer says the assistant must be 'fast and accurate' without thresholds, and not where a government customer requires a compliant deployment but cannot identify data classifications.
Question 399
A sales team expects full autonomy while risk officers expect a drafting assistant. Which action addresses the root concern without unnecessary scope?
- In the project scope, facilitate a decision-rights workshop and record who can approve external actions.
- Within the rollout design, conduct role-specific discovery to document tasks, exceptions, handoffs and pain points.
- For the next architecture revision, rank use cases by feasibility, measurable value, data readiness and risk exposure.
- At the system boundary, run a data-flow and policy discovery workshop before selecting hosting or connectors.
Explanation
Correct Answer
A:Conflicting expectations about authority must be resolved before the architecture can be responsibly designed; the implementation should produce evidence that a signed RACI identifies final decision makers and required human gates.
Incorrect Answers
B:Frontline constraints often differ from executive assumptions and determine whether automation is viable; the trigger would be that a sponsor demands an agent demo before anyone has interviewed frontline caseworkers, whereas the current issue is that a sales team expects full autonomy while risk officers expect a drafting assistant.
C:Portfolio prioritization prevents spending scarce capacity on the loudest unsupported request; the design helps when a company has eight AI use cases but one quarter of funding and no shared prioritization, whereas the release criterion addresses a sales team expects full autonomy while risk officers expect a drafting assistant.
D:Deployment choices depend on information categories and regulatory obligations not yet established; that fits cases where a government customer requires a compliant deployment but cannot identify data classifications, while the required safeguard concerns a sales team expects full autonomy while risk officers expect a drafting assistant.
Question 400
A pilot group consists only of enthusiastic power users although the rollout targets all staff. What is the most appropriate initial architecture decision?
- For this workflow, run a data-flow and policy discovery workshop before selecting hosting or connectors.
- For the affected service, rank use cases by feasibility, measurable value, data readiness and risk exposure.
- In the project scope, conduct role-specific discovery to document tasks, exceptions, handoffs and pain points.
- Within the rollout design, recruit representative users and edge-case owners for structured interviews and tests.
Explanation
Correct Answer
D:Biased discovery sampling can hide accessibility, workflow and training constraints; the intended outcome is that research participants cover major roles, skills and exception-heavy operating conditions.
Incorrect Answers
A:Deployment choices depend on information categories and regulatory obligations not yet established; this applies when a government customer requires a compliant deployment but cannot identify data classifications, but here a pilot group consists only of enthusiastic power users although the rollout targets all staff.
B:Portfolio prioritization prevents spending scarce capacity on the loudest unsupported request; the alternative helps if a company has eight AI use cases but one quarter of funding and no shared prioritization, instead of when a pilot group consists only of enthusiastic power users although the rollout targets all staff.
C:Frontline constraints often differ from executive assumptions and determine whether automation is viable; it is justified if a sponsor demands an agent demo before anyone has interviewed frontline caseworkers, not the case in which a pilot group consists only of enthusiastic power users although the rollout targets all staff.