ABSTRACT
There are a variety of real-world domains where “analysis” of incrementally arriving perceptual input requires the simultaneous solution of a set of hard problems. We claim that basically the same set of interacting problems appears in each of these domains. While people can solve the problems in each of these domains, no automated system is close to solving the full set of interacting problems in any of these domains. The purpose of this paper is to state the set of interacting problems that needs to be solved, and present a cognitively motivated framework in which a solution is possible. Our interest is in a solution to the full set of interacting problems. We are not interested in modeling a particular statistical correlation in a data set that appears under certain conditions. An automated solution would be best, but a man-in-the-loop solution that greatly increases a person’s efficiency an/or accuracy would also be acceptable.
