From Magritte to MacGyver : creativity and uncertainty reduction

Abstract

Creativity is defined as the set of reciprocal cognitive and behavioural processes, respectively creative thought and creative action, involved in producing output that is novel, original, and valuable. Drawing on active inference and predictive processing, this paper examines how the drive to reduce uncertainty subserves creativity. Through creative thought, an agent hypothesizes about a configuration of objects or ideas that, by virtue of its novelty and originality, does not yet exist. As this configuration is not yet extant, there are minimal sources of sensory signals to test whether it will actually exhibit the physical and creative properties the agent predicts it to have. Uncertainty regarding the configuration thus arises. To minimize uncertainty and test predictions about the real-world features of the configuration, the agent must create a corresponding sensory landscape, thereby resulting in creative action. Variations in this process result in some commonly observed patterns of creative action, such as leaving output uncompleted, jumping from one project to another, repeated yet varied treatment of the same subject, producing work based on or derived from the output of another creator, and prolonged, continuous work on a project.