Dr. Maxion's research covers several areas of computer science, including development and evaluation of highly reliable systems (hardware, software, operator interfaces and help/documentation), machine-based concept learning (for anomaly detection, user profiling, masquerade detection), and human-computer interfaces. He is developing dependable systems for automated detection, diagnosis and remediation of faulty or unanticipated events in many domains -- international banking, telecommunications networks, vendor help systems, semiconductor fabrication, information warfare and others.
One type of dependable-system application is diagnosis of faults (or other conditions) in new or evolving situations. For example, diagnosis in real-time networks or semiconductor fabrication or information warfare is difficult because, due to continuous environmental changes, there exists no stable model of acceptable performance against which observed behaviors can be judged. Similarly, a robot, or other autonomous computational organism, finding itself in unfamiliar circumstances, must determine with confidence which elements of its environment are normal, and then classify and respond correctly to novel, anomalous or special events. A major research goal is to model the cognitive processes that make such tasks seem so easy for humans.
A final example regards the time consumed in groping through a user
interface. If it takes too long, or if the wrong path is taken, the
result can be catastrophic in a mission-critical system.
If systems are to be depended on, they
must be designed to accommodate the strengths and weaknesses of the
human as a system component.