Autonomously Managing Competing Objectives to Improve the Creation and Curation of Artifacts

DARCI (Digital ARtist Communicating Intention) is a creative system that we are developing to explore the bounds of computational creativity within the domain of visual art. As with many creative systems, as we increase the autonomy of DARCI, the quality of the artifacts it creates and then curates decreases—a phenomenon Colton and Wiggins have termed the latent heat effect. We present two new metrics that DARCI uses to evolve and curate renderings of images that convey target adjectives without completely obfuscating the original image. We show how we balance the two metrics and then explore various ways of combining them to autonomously yield images that arguably succeed at this task.