PATTERNS OF CREATIVE DESIGN: PREDICTING IDEATION FROM PROBLEM FORMULATION

The main objective of our research is to understand the role of problem formulation in creative design ideation. To that end, we have used the web-based testbed of the Problem Map (P-map) computational framework which represents designers' problem formulation in terms of a series of state models, where each state consists of six types of entities in addition to relations within and between different entity types. We gave two design problems to twenty five graduate students in an advanced product design course. We collected their problem formulation data in the P-map testbed and their ideation data through concept sketches. We conducted correlation analysis between variables extracted from the P-maps, and the ideation metrics. We also built regression models for each of the ideation metrics as the dependent variable, and the P-map variables as the independent variables. We used the data from the first problem to predict the ideation scores for the second problem. The predicted results were compared to the actual outcome reported by an independent panel of judges. Models of variety, average and max quality had more accurate predictions while average novelty, average and max quality had statistically more reliable models.