SociAl Agents For Communicative Tasks

Agents which interact with humans on behalf of humans need a model of human social interaction in order to act appropriately. This paper is a brief description of such a model used by an agent to generate e-mail messages. 1. The problem There is a great deal of research in AI on multi-agent systems [Tambe et al. 1995; Etzioni, Lesh, & Segal 1993]. But there seems to be a growing need for agents which can interact not only with other agents but with humans as well. For example, the rapidly growing popularity of email has already made it difficult for some people to respond appropriately to all the messages they receive. Reading a user’s e-mail and generating appropriate responses is a task seemingly wellsuited to intelligent, independent agents. Yet, if an agent is to act as an intermediary between people, the agent must have knowledge of how humans interact with each other in order for the agent’s messages to be socially appropriate and productive. Current research in multi-agent systems and in discourse processing does not offer a model of human social interaction. This paper presents a brief overview of such a model and of an agent (a "proxy") which uses the model generate socially appropriate e-mail messages. In order to clarify what I mean by "socially appropriate" messages and show how a proxy could fail to meet that criterion, I will start with an example of a message which is socially appropriate. The context for the example is that one has been invited to speak before an academic department about one’s research, but the scheduled time of the speech conflicts with a previous commitment, so one must decline. The note in Figure 1 is an appropriate response to such an invitation because it shows a fitting level of *This work was supported in part by the Advanced Research Projects Agency of the Department of Defense, monitored by the Office of Naval Research under contract N00014-93-1-1212. The Institute for the Learning Sciences was founded in 1989 with the support of Andersen Consulting, part of the Arthur Andersen Worldwide Organization. gratitude for the offer, in addition to communicating the basic message that the writer cannot accept the offer. Without the knowledge that a human should show gratitude for offers he or she receives, an agent acting on behalf of such a person might respond merely, "I decline your offer." Upon receiving such a curt response, an inviter would probably feel insulted that the invitee did not value the invitation. One cannot entirely avoid the negative impact of declining, but there are a number of things one can say to lessen the negative impact, such as thanks, apologies, and excuses. Furthermore, an inviter expects such niceties to be observed, so even if the overtures do not change his mind, he is likely to believe that the invitee was acting in good faith. Thus, an agent which interacts with humans on behalf of a human must have some knowledge of appropriate social communication (even if that knowledge is implicit in a procedure or data structure, e.g. a decision tree of response-letter templates indexed by the category of message received).