The Delusional Hedge Algorithm as a Model of Human Learning from Diverse Opinions