I think attributing the accuracy to debates is a rather hasty conclusion. I would love to see an experiment where the individual participants rated their confid

I think attributing the accuracy to debates is a rather hasty conclusion. I would love to see an experiment where the individual participants rated their confidence from 1 to 10, and only the highest confidence answer in each group was taken. My hunch is this would perform just as well (if not better) than post-discussion.