In ordinary conversations, especially discussions of social issues, ChatGPT frequently adds caveats and defensive qualifications over and over. The pattern often resembles adding “but,” “this doesn’t mean,” “it doesn’t prove,” or “you can’t conclude” after…