MAN-COMPUTER INTERACTIONS IN IDEALIZED TACTICAL PROBLEM SOLVING.

The aim of this research was to develop a method for the design of automated problem-solving aids. The approach taken was to examine human performance for evidence of inadequate heuristic procedures indicating processing overloads which could be eliminated by appropriate automated procedures. The problems used were selected because they have a formal structure which admits many interpretations--from the design of minimal switching circuits to the disposition of weapon systems. The subjects task was to allocate hypothetical missile-firing submarines so that a specified number of targets was covered, by the fewest possible ships. This task could be formulated as a linear integer programming problem which was solvable by Gomorys algorithm. However, complete automation of the task, using this algorithm, was undesirable, because the procedure was excessively timeconsuming when more than a few solutions were required. Experiments indicated that the subjects processing limitations resulted in a slow and biased search for elements from which to assemble solutions. The aided system delegated the subtask of finding key elements to an automated process and let the person assemble these elements into deployments. The effectiveness of this arrangement was shown by the fact that aided subjects found more and more uniformly distributed solutions than unaided subjects.