Intelligent Interfaces for Mission-critical Systems

Abstract

Although expert Systems continue to develop at a rapid pace, human experts still make most of the important decisions in our society. When these experts make a mistake, society is often quick to cry foul. Experimental work in the behavioral sciences, however, has shown that humans often commit errors for the same reasons that make them superior to artificial Systems to begin with, i.e., because they have acquired a wealth of problem-solving heuristics and intuitions that work well in general but are rarely optimal in a specific problem context. In this chapter, we describe some recent work on the development of intelligent interfaces that can anticipate the circumstances of human errors arising from naturally formed heuristics, or cognitive illusions, and provide appropriate feedback and training to eliminate them. The approach involves identifying the conditional probabilities relating objective events to subjective impressions of event probabilities and determining whether the decision making strategy implied by these conditional probabilities is suboptimal.