Principal Investigator Howard Shrobe
We are developing a discussion capture and facilitation tool that will let a group note taker capture ideas and decisions on an electronic whiteboard. These ideas will be saved along with a full audio/video capture of the event into the participants’ personal, pervasive data bins.
Electronic whiteboards have been used often for capturing meeting notes and discussions. Researchers at Xerox Parc developed the Liveboard, the first electronic whiteboard. They subsequently built a note-taking tool from their Tivoli gesture recognition tool. This system allows free-form sketches to be grouped and manipulated using different pen-based gestures. It also indexes into an audio record of the meeting. This system is more free form than the eFacilitator, however, so it is not able to extract the same level of semantics from the meeting notes. The idea of a personal agent was articulated by Pattie Maes, in 1994. In her vision, agents work autonomously on behalf of their owner. They learn over time from the user’s corrections and clarifications and through interactions with other agents. The purpose of the agent is to aid users in getting information and performing computational tasks, potentially using other agents. More recently, the Haystack project at MIT has focused the need for agents that operate entirely on a user’s personal data. The ubiquitous computing literature has always stressed the need for an everywhere-available personalized information sphere.
The importance of this application lies in its incorporation of all of the major elements of pervasive computing: sensors, handheld computers, location awareness, multi-agent systems, and anytime/anywhere personalized information access. By providing an application that people actually use, we hope to discover the directions that our pervasive infrastructure must take. It will also demonstrate a vision whereby users are able to get all of the benefits of an intelligent environment watching them while largely maintaining their privacy. While many environments may record a user’s actions, this data will be carefully guarded and distributed only to the people who can benefit by it.
The diversity of components in this application presents many opportunities for additional work. Specifically, a user’s activities can be captured in situations other than meetings. Her activity calendar, emails, current work, and her day-to-day life captured on video are all candidates for things to put in the common data store. Another area of development is the query interface itself. The retrieval agent should adapt over time both with and without direct user supervision. The final extension area to this work is the agent to agent communication. This should also adapt over time, so that the user can ask a question of its own agent, which will be able to find another agent that has this knowledge if it does not know it itself.