Ph.D. Thesis Proposal

Since the inception of AI, great strides have been made towards achieving the goal of extending natural language conversation as a medium of interaction with machines. Today, we find many Conversational Agents (CAs) situated in various aspects of our everyday life. Most of the existing work on CAs has focused on agents that support a single user in each interactive session. On the other hand, people organize themselves in groups such as teams of co-workers, family and network of friends. With the massadoption of Internet based communication technologies for group interaction, there is an unprecedented opportunity for CAs to support interactive situations involving multiple human participants. Support provided by these CAs can make the functioning of some of these groups more efficient, enjoyable and rewarding to the participants. In our work on supporting Collaborative Learning activities involving two or more students, we have identified two problems that must be addressed in order to embed effective CAs in Multi-Party Interactive situations (MPIS). The first problem highlights the technical challenges involving the development of CAs in MPIS. Existing approaches for modeling agent behavior use assumptions and formalisms that cannot address situations that much more common in MPIS compared to interactive situations involving only one user. To address this problem, we have developed Basilica, an event-driven software architecture that alleviates these problems by providing rich representation capability and flexibility to address complex interaction dynamics. The second problem involves the choice of appropriate agent behavior. In MPIS, agents must compete with human participants to get the participants to pay attention to its contributions. As the number of participants in the interactive situation increases, it becomes increasingly important for agents to demonstrate effective communication skills. Empirical research in small group communication has observed two fundamental processes in group interaction i.e. Instrumental (TaskConversational Agents in Multi-Party Interactive Situations 4 related) and Expressive (Social-Emotional). Expressive aspects of interaction have been largely ignored in the work on Conversational Agents. Based on Social-Emotional interaction categories, we have developed and implemented a set of social interaction strategies. A series of experiments described here investigates the effectiveness and appropriate use of these strategies. Besides providing a set of interactive behaviors useful for MPIS, this work suggests a methodology for making CAs better communicators using existing research in small group communication.