Towards Incorporating Technology to Enhance the Stereotype Production Method in Warning Symbol Design

摘要

Warning symbols must communicate effectively to a wide range of people. To achieve this goal, two major challenges must be overcome. First, designers must acquire user input while minimizing any unnecessary design input needed from non-users. Second, designers must develop symbols with a high likelihood of communicating effectively to a population of users that may be diverse in culture, language, and country of origin. Purpose: We evaluated clustering algorithms as a means to enhance the judgment required by designers using the stereotype production method for symbol development. Methods: Sixty-six symbol sketches, 35 from U.S. participants and 31 from Indian participants, were evaluated for the warning referent "Hot Exhaust." A panel of three certified safety professionals semantically annotated the sketches and developed a frequency matrix that associated the presence of graphical attributes with each sketch. Direct clustering and Simple K-Means clustering were performed on the matrix. Results: Mathematical clustering was successful in identifying population stereotypes. The Simple K-Means analysis of the combined nationality matrix produced five clusters, each characterized by a hypothetical centroid symbol analogous to the population stereotypes of the participant group. Only three of the original 35 attributes were contained among these centroids, meaning that the primary differentiators of the population stereotypes were these three primary attributes. Direct clustering found the same three primary attributes - "pipe/stack," "emission lines," and "flame." Further, clustering the nationalities separately revealed that some attributes were universal between the two nationalities, while others seemed to have a culture or country-of-origin sensitivity. Conclusion: Clustering can be used by designers to group sketches into similar families and to identify the attributes of most interest for final symbols. Furthermore, some attributes appear to be "recessive" while others appear to be "dominant" with regard to culture.