Abstract

Research in artificial intelligence and autonomous agents envisions that future robots will accompany humans in their daily lives. The aim is to provide support not only for routine, challenging, or dangerous tasks, but also to improve quality of life through personal assistance and coaching. In order to allow artificial agents to communicate sensibly and to participate in human society, it is important to equip them with the ability to perceive and appreciate aesthetic features of design in a human-like manner. The present study investigates how methods from anthropocentric biocybernetic computing (ABC) can be assembled in an intelligent control module for architectural design evaluation. Central to the system is an abstract model of aesthetic experience, which is established through statistical learning. For the experiments, a database of images of house façades is employed. The learning algorithm extracts line distributions, which characterise façade design, and represents them abstractly in the form of a non-linear manifold. Each point on the manifold corresponds to one façade. The proposed module includes two additional affective perceptual pathways, which are implemented using paradigms that are believed to reflect responses of the human emotional system. One paradigm involves concepts of facial expression recognition, and the other is based on calculating the fractal dimension of the skyline of cityscapes. Future applicability of the proposed system for design evaluation will rely on suitable data preparation and calibration of the associated algorithms using test subjects. The article describes characteristic details of the system’s architecture and discusses whether it would be able to acquire the level of sophistication required to provide aesthetic judgment that is convincing for humans.