Toward a Computational Model of "Context"

According to a well-known statement by Donald Knuth, “Science is what we understand well enough to explain to a computer. Art is everything else we do.” If we use Knuth’s criterion, we must probably conclude that contemporary investigations of the ‘context’ of a social activity – as offered in fields such as linguistic pragmatics and social psychology – are still art rather than science (Reich 2010). The blunt truth is that social scientists have tended to treat context as a residual category that encompasses ‘everything else’ with regard to the object of interest (a communicative message, a social situation, and so on). The few explicit attempts to theorize context that do exist (e.g., Gumperz 1992; Sperber and Wilson 1995, chapter 3; Akman 2000; Dijk 2006) are general and exploratory, hence only marginally helpful for the purpose of specifying a precise, operational, implementable model. Turning the elusive notion of context into such a model is a necessary task if researchers are to succeed in building artifical agents – be they robotic or virtual – that can interact and collaborate with humans on human terms (Bunt 2000). For instance, how is the robotic agent in figure 1 to distinguish between a pointing gesture that means (a) “Give me the tool” and one that means (b) “There is the tool you need”? The purpose of this paper is to reinvigorate computational research on context by discussing three components of the context of extremely simple communicative acts in one specific social domain: communicatively coordinated collaboration (CCC). With this term I refer to a social situation in which two or more agents cooperate on a task by way of using overtly intentional communicative acts (which include not only speech acts but also nonverbal acts, indeed any behavior which is meant to influence a co-agent in an overt manner; Reich 2011). CCC can be contrasted with tacitly coordinated collaboration (TCC), which involves agents cooperating by watching and responding to each other’s pursuits without overtly intentional signals. Humans in all