| While many areas of man-machine interaction have well-developed performance measures which are used to evaluate the effectiveness of a certain tool for a task performed by a human, computer-aided architectural design (CAAD) does not. Therefore, the goal of this study was to develop a performance methodology which will be useful for evaluating human performance for different types of tasks on a given system and across different levels of complexity within a single task.; To meet the above goals, performance measures that reflect meaningful changes in humans behavior during CAAD tasks were developed. These measures were based on models of human information processing. Two cognitively different architectural tasks were formulated which differed in terms of the stimulus-central processing component-response compatability and the structuredness of their problem spaces. Methods of varying task complexity within each of these tasks were also developed to test the sensitivity of the performance measures across levels of complexity and to introduce variability into the humans design behavior.; Five University of Buffalo Architectural graduate students and one architect participated in the experiment. Each subject was trained to use a CAAD system and then they each performed the two design tasks on the computer (each subject was assigned to one of three levels of complexity).; The data collected included time-stamped interaction records, protocols, and a videotape of the computer screen. Identification of design phases within a task, time lag analysis, graphical summarization techniques, Markov analyses, protocol analysis, and error analysis were performed on the data. The changes in the objective performance measures were shown to relate to changes in cognitive processes underlying human design activities, identified through the information gained from the protocols.; From the developed performance measures task complexity, type of task, and subjective effects on performance could be seen. It was also shown that some measures more directly reflected the computer interaction aspects of the task while other measures reflected the cognitive design activity of the human. Heuristics, chunking strategies and other problem-solving techniques used as a result of human cognitive limitations were identified. The techniques used differed between tasks. |
