ArticlesWINTER 2011 17 O ne of the most promising concepts that emerged in the last two decades of computer science research is embodied interaction -representing computational intelligence in and offering interaction through physical or simulated ''bodies." Such representations vary from physical artifacts (such as tangible interfaces) to biological forms (such as humanlike agents and robots) and offer templates for understanding and interacting with complex computational systems (Ullmer and Ishii 2000, Cassell 2001, Breazeal 2003. Physically embodied humanlike attributes provide particularly rich representations with which people are intimately familiar, and systems that draw on these representations promise significant social and task outcomes in key application domains from education (Kanda et al. 2004) to collaboration (Bluethmann et al. 2003) to rehabilitation (Dautenhahn and Werry 2004).Embodied interaction offers not only a familiar template for making sense of and interacting with computational systems but also an opportunity to achieve the most effective forms of human interaction for achieving positive social and task outcomes in such domains. How do the most effective teachers improve student learning? How do the most effective speakers capture their audiences? How do the most effective personal coaches motivate people to achieve their goals? Research in human communication has shown that brief observations of the bodily, facial, speech, and vocal cues that people display predict these outcomes (Ambady and Rosenthal 1992). Seemingly subtle differences in such cues of interaction partners shape outcomes such as perceptions of their attitudes (Mehrabian 1967), the persuasiveness of their messages (Segrin 1993), and their performance in collaborative work (Burgoon et al. 2002).This tight coupling between communicative cues and key social, cognitive, and task outcomes opens up a space for designing effective interactions for robots to elicit particular positive