The effects of problem structure and dialogue type on the performance of the man/machine interface

Abstract

This research project is intended to provide guidance for software designers who must decide on a dialogue style for interactive problem solving support for new or infrequent users. In addition, it examines the relative performance of these dialogues in environments with a varying amount of problem structure. The three dialogue styles that are recommended by various writers in the field of information systems are: (1) menu driven, (2) question/response, and (3) form filling. Real world managers using systems based on these dialogue styles, are asked to solve one of two problems which are different in terms of complexity and the number of parameters used to represent the problem. All other aspects of the problem are identical. The problem's structure affects the profit performance of the subjects, but within a given level of problem structure they perform equally well (on average) regardless of the type of dialogue style used to carry out the interaction. The syntactical performance of the subjects is unaffected by any of the treatment manipulations, but the percentage of incorrect inputs is so small that this response is, in hindsight, not an important factor. We conclude that if an interface is properly designed it probably does not matter which dialogue style is used in some problem settings. It is also clear that adding problem parameters reduces the ability of the problem solver to determine the best solution.