Estimation of discrepancy of color qualia using Kullback-Leibler divergence

Abstract

Qualia have traditionally been considered difficult to measure objectively, but with the recent spread of fMRI (functional Magnetic Resonance Imaging) and other techniques, various experimental efforts have been made. In this paper, focusing on the qualia for color, we created 6 colors with different RGB values for reference colors of RED, light GREEN, BLUE, YELLOW, and PURPLE, and showed them to 306 subjects. For example, for RED and 5 generated colors, we asked them, “Choose a color that you feel is RED,” and asked them to choose. A probability density function was defined for each of the five generated colors and the reference color, which is the primary color of RED, light GREEN, BLUE, YELLOW, and PURPLE, and the Kullback-Leibler divergence between the probability density function of the reference color and the generated color was calculated, the relationship between the number of samples of the selected color and the Kullback-Leibler divergence was obtained, and the difference in color sensation-qualia was calculated accordingly. As a result, it was confirmed that the larger the distance of the Kullback-Leibler divergence, the smaller the number of samples, but that the distribution shape in which the number of samples decreased for each color differed greatly. This suggests that if we see a color such as RED to PURPLE, we are randomly choosing a color that “feels."