I bet you are right. And, it would be really fascinating to see how "right" the AI wanted to be with the candidate answers. Thinking about that has me go down a rabbit hole, wondering how my learning would be improved or impeded by the "right" wrong answers.