To better understand the failures of machine learning systems, it's important to ask what the machine was thinking in those edge cases. In robotics, naming or gendering robots can lead to personification and a construction of a life story, which may not be ideal.