Trace predictions through staff outreach: Before approval, which evidence best determines whether review must
Because outputs trigger human actions and applicant responses, review should cover the complete sociotechnical workflow.
The question
A research administration office plans a model that predicts which proposals may need budget clarification. Staff will prioritize outreach, and applicants may revise submissions after those contacts. Before approval, which evidence best determines whether review must cover only the model or the wider system?
Preparing for AIGP? Take the free 5-min readiness quiz →
- Review model architecture, feature documentation, and rankings of feature importance.Technical documentation can clarify model behavior, but it does not capture the human workflow and applicant effects created by deployment.
- Evaluate prediction accuracy on previously submitted proposals.Accuracy evaluates the model’s predictions but does not examine staff outreach, applicant responses, or resulting decisions.
- Survey administrators about whether they understand the model’s purpose, limitations, and escalation process.Administrator understanding is relevant oversight evidence, but alone it omits actual operational behavior and effects on applicants and decisions.
- Trace predictions through staff outreach, applicant responses, and resulting proposal decisions. ✓The deployed sociotechnical system includes the model, staff actions, applicant responses, and downstream decisions, so evidence must cover that complete workflow.
The trap
Follow an output into human actions, affected people, feedback, and downstream consequences. How to remember it
Because outputs trigger human actions and applicant responses, review should cover the complete sociotechnical workflow.
How many of these would you get right?
One of 1581 AIGP questions on Certsqill. Take a free five-minute check and see your score per domain — not one number, but which section to open tonight.
Test your AIGP readiness — freeMore Understanding the Foundations of AI Governance questions
- Classify the output as a forecast or generated instruction: Before selecting evaluation and governance →
- Calibration results plus agent override performance: Which evidence is most decision-relevant? →
- Testing counselor intervention rates and error detection: Before approval, which evidence best resolves the →
- All 337 Understanding the Foundations of AI Governance questions →
Part of the Certsqill AIGP question bank · Understanding the Foundations of AI Governance ·
Every answer, right and wrong, comes with its own explanation.
