Modern AI image classifiers have made impressive advances in recent years,\nbut their performance often appears strange or violates expectations of users.\nThis suggests humans engage in cognitive anthropomorphism: expecting AI to have\nthe same nature as human intelligence. This mismatch presents an obstacle to\nappropriate human-AI interaction. To delineate this mismatch, I examine known\nproperties of human classification, in comparison to image classifier systems.\nBased on this examination, I offer three strategies for system design that can\naddress the mismatch between human and AI classification: explainable AI, novel\nmethods for training users, and new algorithms that match human cognition.\n