How People Talk with Robots: Designing Dialogue to Reduce User Uncertainty

Abstract

WINTER 2011 31 Verbal interaction with robots is not characterized by a set of well-defined properties; what users actually say when faced with a robot, especially if they are not engineers or computer scientists, can hardly be predicted. In this article I argue that the peculiarities of human-robot dialogue are best understood as users’ strategies to deal with what they understand the challenges of the situation to consist in (Fischer 2006a). That is, users interact with artificial communication partners on the basis of what they consider to be potentially problematic, what the task comprises, what the robot can understand, and so on, that is, what they consider its affordances to be (Gibson 1977). The variation observable between users is then due to different concepts of what the affordances of the system really are. Although this observation seems trivial, there are in fact several other suggestions for what human-robot interaction might be shaped by. The most prominent is certainly the idea that users transfer mindlessly from social interactions between humans to interactions with computers and robots (for example, Nass and Moon [2000], Nass and Brave [2005]). In particular, Nass and colleagues have investigated a broad range of social behaviors, with the result that people were found to react to a computer’s behavior as they do to other humans’ behavior (Fogg and Nass 1997), or that they transfer to the agent human characteristics, such as intentionality (Ju and Takayama 2009), ethnicity (Pratt et al. 2007) or gender (Nass and Brave 2005). Nass proposes that the reasons for this transfer, which has also been called media equation (Reeves and Nass 1996), is mindlessness, an error due to thousands of years of evolution in social environments (Nass 2004). A second proposal, which often remains implicit, is that there is either a particular conventional way of speaking to computers and robots, or that users make use of other conventional varieties, for instance, the way of talking to children. The linguistic term for such a situation-specific variety is register (Krause and Hitzenberger 1992). However, the interpersonal variation observable in the ways in which people talk to robots (see Fi scher 2006a, 2006b, 2011)