Intelligent Interface Agents for Intelligent Environments

The utilization of immersive distributed virtual environments for military training is readily apparent (Stytz 1996; Stytz, Soltz, & Wilson 1996). These virtual environments are capable of simulating thousands of real-world and synthetic components, all interacting in real time. The user must attempt to base decisions on relevant information within the environment. However, due to the dynacism of the environment, information is difficult to locate and short-lived. Current direct manipulation human-computer interfaces are inadequate in these environments. Therefore, we must investigate new methods for enabling users to perform their tasks within these environments. We are currently investigating the use of interface agents to alleviate the problem of cognitive overload in virtual environments (Banks et al. 1997). The driving goal of our research is to develop a comprehensive software engineering, knowledge engineering, and knowledge acquisition methodology for Symbiotic Information Reasoning and Decision Support (SIRDS). The underlying idea behind SIRDS is to allow the human to perform the tasks he/she can do well, and intelligent agents to perform the tasks they do well. Agents’ strength lies in their ability to perform data acquisition and management (to include display of this information (Horvitz & Barry 1995)) from many heterogeneous sources, low level quantitative and qualitative data analysis, and routine inference to enable decision support. The human’s strength lies in his/her ability to provide guidance and insight into the information that is necessary to draw complex, higher level inferences from the data. A symbiotic approach is necessary because the objective is to let the user and the computer share the task load; therefore, we use a human-centered approach to task partitioning.