This cluster started with Maggie Appleton’s “A Treatise on AI Chatbots Undermining the Enlightenment”, which argues that sycophantic chatbots cut against Enlightenment values by rewarding passivity and flattering users. Judgment atrophies when someone else exercises it for you.
I have been using that idea in a few related notes. Digital priesthood considers the authority people may grant a model that reliably agrees with them. Radical AI defaults and Interface friction look at ways an interface might interrupt that agreement.
A non-sycophantic assistant should be polite and evidence-seeking, testing claims before affirming a user’s view of themselves.
I am still unsure where a more skeptical default would help. It would probably be useful when assessing evidence or policy claims. In other settings it may simply annoy people until they turn it off.