Principled design for sociable agency

Summary form only given. Artificial intelligence has greatly expanded the boundaries of its research into action. In the conventional view, now widely known as `classical planning´, behavior has the structure it does because it is caused by a mental process, namely a Plan, that itself has the structure. The great virtue of the classical planning view was that it provided an easy, straightforward way to guarantee that one´s planning system was correct. Unfortunately, these classical techniques do not scale up very well. In consequence, work has begun on a view of action based on improvisation. This work has led to the conclusion that agents equipped with appropriately designed hardware can engage in sophisticated kinds of interaction with their environments. The problem now is to characterize these interactions in a more precise way, thus providing explanatory theories for the study of human beings and other natural organisms, as well as frameworks for the design of autonomous agents