Social robots are robots designed to interact with humans or with each other in ways that approximate human social interaction. It seems clear that one question relevant to the project of designing such robots concerns how humans themselves interact to achieve social understanding. If one turns to psychology, philosophy, or the cognitive sciences in general, one finds two models of social cognition vying for dominance under the heading of theory of mind: theory theory (TT) and simulation theory (ST). It is therefore natural and interesting to ask how a TT design for a social robot would differ from the ST version. A much more critical question is whether either TT or ST provide an adequate explanation of social cognition. There is a growing although still minority consensus that, despite their dominance in the debate about social cognition, neither TT nor ST, nor some hybrid version of these theories, offers an acceptable account of how we encounter and interact with one another. This paper presents a brief review of the theory of mind debate, outlines an alternative theory of social cognition based on an embodied interactive approach, and then tries to draw out a few implications about social robotics.