Preferred Surface Luminances in Offices, by Evolution

Abstract

Forty participants viewed a series of greyscale images of a typical non-daylit, open-plan partitioned office, and rated them for attractiveness. The image was projected onto a screen at realistic luminances and 54 percentof full size. The images in the series were geometrically identical, but the luminances of important surfaces were independently manipulated. Initially, the combinations of luminances were random, but as the session continued, a genetic algorithm was used to generate new images that generally retained features of prior images that were rated most highly. As a result, the images presented converged on an individual's preferred combination of luminances. The results demonstrated that this technique was effective in reaching a participant's preferred combination of luminances. There were significant differences in room appearance ratings of the most attractive image compared to other images, and the differences were in the expected direction. Factor analysis of ratings of the most attractive images revealed a factor structure with some similarity to that obtained when people rated real office spaces. Furthermore, preferred luminances were similar to those chosen by people in real settings, as was the variation in preferences between individuals. Finally, subjective ratings of brightness, uniformity and attractiveness were significantly related to luminances in the image.