next up previous
Next: System Overview Up: Tracking Conversational Context for Previous: Tracking Conversational Context for

Introduction - The Computer as a Mediator

The recent progress of the Human-Computer-Interaction community has produced a variety of intelligent interfaces to assist humans in their interaction with a computer. In the process, various supporting technologies are called upon (tangible interfaces, audio-visual modalities, etc.) to produce a more natural look-and-feel to the computer. However, the emphasis has been primarily on improving the interaction between a single human and a computer.

We propose an alternative paradigm where the computer acts as an external third party. It observes several humans interacting with each other and assists them as an external mediator. Our specific application is a group meeting where several individuals are actively involved in a discussion. The computer is no longer the focus of attention and must autonomously track the conversation, provide feedback and augment the meeting. As 'external mediator' the machine must have some situational awareness or understand the conversational context such that it intelligently enhances and assists the overall interaction. A human facilitator has a similar task. He can be assistive without being an expert, having background knowledge and understanding exact details of the group meeting.

Another constraint is that a machine should be autonomous such that it doesn't encumber the individuals, restricting their conversation by requiring direct input and manipulation. Therefore, the computer must participate in a non-obtrusive manner, using audio-visual or passive modalities that allow the participants to continue carrying out a conversation without browsing menus, using a GUI or getting slowed down by the 'mediator'. Since it is not expected to directly replace one of the participants (i.e. as an automated bank teller), the computer does not have to carry the conversation itself1. It instead stimulates it further (implicitly or explicitly) and provides relevant information when appropriate contextual cues arise from the conversation between the humans.



 
next up previous
Next: System Overview Up: Tracking Conversational Context for Previous: Tracking Conversational Context for
Tony Jebara
2000-02-24