[ascorbic]

Pushing back when the user is wrong is a very desirable feature, whatever the mental health of the the user. I can't think of any scenario when it's better for an LLM to incorrectly tell the user they're right, instead of pushing back.