Question 35
Domain 2: Fairness, Bias, and Societal ImpactA team is designing an AI content recommendation system that could influence users' beliefs and behavior. Which safeguard is the best choice to reduce harmful societal impact while still allowing the system to operate?
Correct answer: B
Explanation
Content systems that can shape beliefs or behavior should include safeguards specifically designed to limit harmful influence while allowing beneficial use. The appropriate control is one that directly addresses the system's power to affect users, rather than only improving performance or engagement. — Safeguards for content systems that influence beliefs or behavior.
Why each option is right or wrong
A. Increase the frequency of personalized recommendations to improve user retention
Improving retention does not address the need for safeguards on systems that influence beliefs or behavior.
B. Implement safeguards that limit harmful effects from content influencing beliefs or behavior
The source material specifically recommends safeguards for content systems that influence beliefs or behavior. In this scenario, the recommendation engine has that influence, so adding safeguards targeted at harmful societal effects is the directly supported response.
C. Expand the range of persuasive content so users receive more viewpoint-shaping material
Adding more viewpoint-shaping content increases influence rather than introducing protections against harm.
D. Optimize the model only for relevance so social effects are outside the system scope
Societal effects remain relevant when content systems influence beliefs or behavior.