Misalignment with the intended objective combined: Which combination of AI risks is most clearly illustrated?
Optimizing a watch-time proxy is misalignment; scale-emergent, hard-to-fix harm is complexity and scalability.
The question
A platform's recommendation AI maximizes watch time and, over months, amplifies divisive content, harms some users' well-being, and proves hard to correct because effects emerge only at population scale. Which combination of AI risks is most clearly illustrated?
Preparing for AIGP? Take the free 5-min readiness quiz →
- Data-dependency and security risk, since compromised training data caused the harmful content amplification observedPlausible but wrong: nothing indicates poisoned or compromised data; the harm flows from the optimization target at scale.
- Misalignment with the intended objective combined with complexity and scalability harms that surface only at scale ✓Correct: optimizing watch-time as a proxy is misalignment, and scale-emergent, hard-to-correct harm is complexity and scalability.
- Opacity and autonomy risk alone, since the system acted independently and its internal logic was wholly undocumentedAlmost right: opacity may contribute, but the defining issues are objective misalignment and scale-driven harm, not autonomy alone.
- Bias risk exclusively, since the harms stem entirely from unequal treatment of protected demographic user groupsPlausible but wrong: the scenario describes objective and scale harms, not systematically unfair treatment of a protected group.
The trap
Reducing a compound objective-and-scale failure to a single cause such as bias or data compromise. How to remember it
Optimizing a watch-time proxy is misalignment; scale-emergent, hard-to-fix harm is complexity and scalability.
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
- A generative, general-purpose model, since it produces new: Which characterization is most precise? →
- Autonomy, speed and scale, and opacity together, so harms: Which characteristics most justify a comprehensive, →
- Add explainability tooling and bias testing so: Which approach best honors the responsible-AI principles in →
- 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.
