The objective of this research is to evaluate the similarity of design problems used in research experiments based on their representation. Design problems form an essential component of experiments in creativity research. However, the formulation of a design problem for an experiment is still a challenging task. An opportunity for the design community is leverage existing design problems to reduce a source of difference between experiments and increase the validity of new method and tools. To answer this question, fifty-five design problems, published over the past fifteen years, were collected and analyzed using two approaches. First, the information content of the design problems was assessed based on five structural elements. A protocol study was developed with four independent evaluators, who were asked to identify the five elements in a set of design problems. A high correlation was observed for the characteristics: goal of a problem, functional requirements, non-functional requirement and reference to an existing product. A low correlation was observed for the characteristics information about end user. The results suggest that this approach could serve as a useful tool in comparing and selecting problems with a desired number of elements. In the second approach, latent semantic analysis was used to compare the design problems. Latent semantic analysis enables design problem similarity to be computed prior to solving the problem. However, this method cannot identify the nuances in problem representations, which could potentially affect solution development. The paper concludes with an appeal to research community of using similar or benchmarked design problems, and postulate other means to enable comparison or development of problems in creativity research.

This content is only available via PDF.
You do not currently have access to this content.