Question 27
IIIA content recommendation system learns from user clicks and tends to show increasingly narrow content to maximize engagement. Which risk is most relevant?
Correct answer: B
Explanation
Feedback loops arise when a system’s outputs shape future inputs, so showing more of what users already click can “reinforc[e] existing preferences.” This can create “filter bubbles,” where increasingly narrow recommendations “potentially marginaliz[e] diverse or minority content” by limiting exposure to different viewpoints.
Why each option is right or wrong
A. The system might recommend content in languages it does not support.
Language support is a localization or model-capability issue, not the main engagement-feedback risk described.
B. Feedback loops can lead to filter bubbles, reinforcing existing preferences and potentially marginalizing diverse or minority content.
The relevant risk is the self-reinforcing feedback loop: if the recommender optimizes on clicks, its outputs become the next round of training/interaction data, so the model keeps amplifying the same high-engagement items rather than correcting for diversity. In practice, this produces a filter bubble effect, where exposure narrows over time and less popular or minority content is systematically under-served, even though it may be valuable or relevant.
C. The system may run too quickly and overwhelm users with recommendations.
Recommendation speed affects user experience, but not the core bias-amplification problem from repeated click optimization.
D. The system will inevitably generate synthetic user profiles.
Synthetic profiles are not inevitable; the described risk is reinforcement of observed preferences.