Artificial Aesthetics and Aesthetic Machine Attention

Abstract

The aesthetics of artificial intelligence is often viewed in relation to the qualities of their generated expressions. However, aesthetics could have a broader role in developing machine perception. One of the main areas of expertise in aesthetics is the understanding of feature-based information, which involves how the aesthetics of sensory features can cause affective changes in the perceiver, and the other way around – how affective states can give rise to certain kinds of aesthetic features. This two-way link between aesthetic features and affects is not yet well-established in the interdisciplinary discussion; however, according to perceptual psychology, it fundamentally constructs the human experience. Machine attention is an emerging technique in machine learning that is most often used in tasks like object detection, visual question answering, and language translation. Modern use of technology most often focuses on creating object-based attention through linguistic categories, although the models could also be utilized for nonverbal attention. This paper proposes the following perceptual conditions for aesthetic machine attention: 1) acknowledging that something appears (aesthetic detection); 2) suspension of judgment (aesthetic recognition); and 3) making the incident explicit with expression (aesthetic identification and amplification). These aspects are developed through an interdisciplinary reflection of literature from the fields of aesthetics, perceptual psychology, and machine learning. The paper does not aim to give a general account of aesthetic perception but to expand the interdisciplinary theory of aesthetics and specify the role of aesthetics among other disciplines at the heart of the technological development of the human future. Article received: May 10, 2022; Article accepted: July 15, 2022; Published online: October 15, 2022; Original scholarly paper