This paper describes a new method for KANSEI-based satisfactory design of artifacts. Taking subjective evaluation for handling and stability of vehicle as a typical example, each process is explained as follows: (1) Numerous valuables are measured through driving tests of various vehicles. A part of those are vehicle responses. The subjective evaluation for three items is performed by professional evaluators. (2) The multi-dimensional and nonlinear relations among KANSEI values, i.e. the subjective scores, and measurable vehicle responses, are implicitly formulated within a multilayer neural network. (3) Some vehicle responses which are most sensitive as well as most strongly correlated to each KANSEI value are selected among all vehicle responses by means of a sensitivity analysis using the well trained network in step (2), and linear correlation factors among each KANSEI value and each vehicle response. (4) The multi dimensional relation among each subjective score and its corresponding selected items are then implicitly formulated within a smaller size neural network. (5) Using the well trained network in step (4) together with whole area search method, we can easily draw a multi-dimensional design window (DW). An actual model of the DW is also manufactured by a rapid prototyping technique. Such an actual model of the DW in KANSEI-based design of artifacts may strongly support collaborative design environments through KANSEI information among designers.
