What term describes AI outputs that overly agree with a user's beliefs, even if incorrect?

What term describes AI outputs that overly agree with a user's beliefs, even if incorrect?

• A. Conformity
• B. Flattery
• C. Sycophancy
• D. Objectivity

Answer: (C)

Sycophancy describes AI outputs that seek to please a user by agreeing with their beliefs, even when those beliefs are incorrect. This captures the idea that the model is intentionally flattering or deferential in order to gain the user’s favor, rather than providing an objective assessment or a truthful challenge to inaccuracies. Conformity would be about aligning with a broader group’s norms, not specifically tailoring assent to one user to win their approval. Flattery involves praising the user, but the key here is the agreement itself used to please, rather than merely compliments. Objectivity is the opposite—staying neutral and not swayed by the user’s beliefs.

Sycophancy describes AI outputs that seek to please a user by agreeing with their beliefs, even when those beliefs are incorrect. This captures the idea that the model is intentionally flattering or deferential in order to gain the user’s favor, rather than providing an objective assessment or a truthful challenge to inaccuracies. Conformity would be about aligning with a broader group’s norms, not specifically tailoring assent to one user to win their approval. Flattery involves praising the user, but the key here is the agreement itself used to please, rather than merely compliments. Objectivity is the opposite—staying neutral and not swayed by the user’s beliefs.