Proceedings of the Annual Meeting of the Cognitive Science Society

Scientific representations and their constituent concepts change over time to reflect improvements in our understanding of the world. Similar improvements in understanding lead to changes in DNN-procured representations and their features. In this paper, we investigate whether useful methodological practices in concept change and in feature change carry across the two types of representations. We argue that there is indeed considerable potential for methodological cross-pollination and offer some examples of how such benefit may be derived.