Knowledge structures in human problem-solving: Implications for human-computer interactive tasks

David Chester Gibson, Purdue University

Abstract

Under ideal circumstances, no psychological adaptation is required to bridge the discrepancies between psychological and physical knowledge representations in the human-computer interface. Current philosophies of interface design attempt to bridge the human-computer representational gulf through the use of mental models, providing users with some degree of predictive and explanatory power in understanding their interaction with the system. While this technique has it merits, it does not consider the internal knowledge structure of the user. It is proposed that efficient human-computer problem solving might be facilitated if, in addition to mental model considerations, the knowledge structure of the user was incorporated into the interface design. Based on previous research, it was hypothesized that the problem solving characteristics that serve as the basis for Greeno's typology were manifestations of different configurations of a common underlying knowledge structure. The hypothesis was tested in a series of three experiments. Experiment One investigated the nature of knowledge structures employed by twenty subjects in solving nine problems characteristic of Greeno's typology. Multidimensional scaling (MDS) analysis revealed three common underlying knowledge structure dimensions: specific vs. abstract representation, level of conscious vs. subconscious processing, and level of inductive vs. deductive reasoning. The different configurations obtained for each of three problem types (arrangement, transformation, inducing structure) were consistent with the problem solving representations and processes originally hypothesized by Greeno. An identical knowledge structure was identified in Experiment Two, in which ten subjects solved five relatively complex tasks. Experiment Three tested twenty subjects in a simulation programming task, investigating interface applications of the knowledge structure developed in Experiments One and Two. The results of Experiment Three indicate that in human-computer interactive problem solving, tasks associated with arrangement should employ specific interface representations. In contrast, transformation tasks should utilize abstract interface representations. These results indicate that it is possible to initially design the interface problem representation in a manner which is compatible with the user's own problem solving representations and processes, thereby enhancing human-computer problem solving efficiency.

Degree

Ph.D.

Advisors

Salvendy, Purdue University.

Subject Area

Industrial engineering|Occupational psychology|Computer science

COinS