Its outputs are probabilistic and can differ for similar: Which characteristic of AI most directly explains
The described variability for similar inputs reflects AI's probabilistic (rather than deterministic) nature, which is the characteristic creating the new oversight need.
The question
A bank replaces a fixed rules-based loan-approval script with a machine learning model. Reviewers observe that near-identical applicant profiles occasionally receive slightly different scores across model retrains. Which characteristic of AI most directly explains why this system needs governance the old script did not?
Preparing for AIGP? Take the free 5-min readiness quiz →
- Its dependence on training data means the model mostly reproduces whatever historical patterns were present in the records it learned from.Plausible, because data dependency is a real AI characteristic, but it explains bias inheritance, not the version-to-version score variation described here.
- Its speed and scale let it evaluate far more applications per hour than a manual review team of loan officers ever realistically could.Plausible, since speed and scale are governance-relevant, but throughput is not what produces differing scores for the same profile across retrains.
- Its outputs are probabilistic and can differ for similar inputs, unlike the deterministic rules it replaced, so results need monitoring. ✓Correct. The shift from fixed, repeatable rules to probabilistic model outputs is exactly what introduces variability and the need for continuous oversight.
- Its opacity makes the internal reasoning path difficult to inspect, even in the cases where the final output value happens to remain stable.Almost right, as opacity is a genuine concern, but the scenario highlights changing outputs rather than an inability to inspect a stable decision path.
The trap
Assuming any surprising AI output must stem from biased data or opacity, rather than from AI's inherently probabilistic behavior. How to remember it
The described variability for similar inputs reflects AI's probabilistic (rather than deterministic) nature, which is the characteristic creating the new oversight need.
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
- Bias risk arising from unrepresentative training data: Which AI risk are they describing? →
- Human-centricity: Which responsible-AI principle does this override requirement most directly operationalize? →
- Define and assign clear roles and responsibilities so each: Which governance action would most directly →
- 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.
