Decision-theoretic reasoning and the human-computer interface: Advances in embedded intelligent agents

Abstract

Advances in the use of automated decision-theoretic reasoning to enhance the human-computer interface will be presented. Following a brief survey of research on Bayesian networks and influence diagrams, principles of embedded decision-theoretic agents will be discussed. Key concepts will be highlighted within the contexts of work on the Vista and Lumiere projects. The Vista project centered on building an intelligent interface to assist NASA Mission Control flight engineers with managing the complexity of information about Space Shuttle propulsion systems in time-critical contexts. Moving from aerospace to personal computing, related work on the Lumiere project at Microsoft Research will be presented. The goal of Lumiere is to develop more responsive interfaces for personal computers.