Abstract
A long-standing question within the robotics community is about the degree of human-likeness robots ought to have when interacting with humans. We explore an unexamined aspect of this problem: how people empathize with robots along the anthropomorphic spectrum. We conducted a web-based experiment (n = 120) that measured how people empathized with four different robots shown to be experiencing mistreatment by humans. Our results indicate that people empathize more strongly with more human-looking robots and less with mechanical looking robots. We also found that a person's general ability to empathize has no predictive value for expressed empathy toward robots.