“Knowledge From Beauty: Applying Expert Systems to Aesthetic Judgement” by Adelson and Franguiadakis
Conference:
Type(s):
Title:
- Knowledge From Beauty: Applying Expert Systems to Aesthetic Judgement
Program Title:
- G-Tech
Organizer(s)/Presenter(s):
Description:
Knowledge From Beauty generates simple 3D vessel forms defined by seven control points which specify a profile. The control points can be restricted to meet, for instance, ergonomic constraints (e.g., diameter easily grasped by an adult hand). The control points are generated randomly within their limits. There are more than six million possible combinations for the profile. Users rote the forms on a seven-point scale. An expert system module within the program uses geometric factors (such as the width of the base and number of “humps” in the profile) to learn about user likes and dislikes. After the program has some experience with the user, it begins to “filter” the forms which are shown, rejecting geometries that the user is predicted to dislike. After 50 to 70 images, the program’s ability to predict preferences usually improves, and the filtering becomes more accurate. In some cases, the program can learn to be almost as accurate as a person’s own consistency. They also show that the geometric factors that most influence preference can vary a good deal between individuals. Knowledge From Beauty originated as an industrial design master’s degree thesis project at the University of Illinois at Chicago’s Design Visualization Laboratory.
