Adjusting the Novelty Thermostat: Courting Creative Success through Judicious Randomness

Though creativity is usually defined as “novel and appropriate,” this is most often understood to mean “as novel as possible, so long as appropriate.” While this definition might be suitably applied to finished products, it is less obviously useful as a guiding value during the act of creation. This research tests this and other definitions by using computer simulations based on Campbell’s “blind variation and selective retention” theory. Introducing a “temperature” parameter to reduce novelty’s importance over time produces results superior to both an even combination of novelty and appropriateness and the prevailing “novel, so long as appropriate” definition. However, choosing the correct temperature adjustment schedule is essential. In this way, the simulations show that human decision processes might plausibly produce the same dynamics as Simulated Annealing, and thus that metaheuristic optimization can be an insightful theoretical guide for creativity researchers. The results also show the feasibility of Campbell’s theory. Finally, they show that if creativity is to be a guiding value while creating, then it is best defined as “appropriately novel, and appropriate.” The overall action of creative systems is to advance the status quo. This is difficult, making successful cases rare. Creativity research began by trying to learn how to reproduce these cases. This required a standard of when an idea or artifact is “creative.” Because exceptional cases are also different from what came before, the defining characteristic is most often novelty. However, not every novel creation is worth having created: it must also be useful, of value, or broadly, “appropriate.” Thus, the most common definition has become “novel and appropriate.” Though appropriateness is necessary, people usually equate the extent of creativity with the extent of novelty. For instance, Runco and Smith (1992) operationalize creativity with a bipolar scale from “common” to “creative” (rather than “rare”). Similarly, Amabile (1996) goes to great length to argue that creativity is different from aesthetics or quality, but not from novelty. This leads to seeing creativity as being “as novel as possible, so long as appropriate.” Working definitions of creativity probably ∗To appear in Proceedings of the AAAI Spring Symposium on Creative Intelligent Systems, Stanford, CA, March 2008 play an active role in the creative process (Amabile 1996; Runco 2003). If creators ultimately want results that are better, and not simply different, it is important to ask whether “novel, so long as appropriate” is a suitable way to view creativity. Drawing inspiration from the field of metaheuristic optimization (Blum & Roli 2003) and Campbell’s (1960) “blind variation and selective retention” model of the creative process, this paper uses computer simulations to test how well different ways of understanding “novel and appropriate” impact final outcomes. These simulations show that the standard definitions of creativity are incomplete. They also demonstrate that psychological processes can plausibly produce the same dynamics as metaheuristic optimization algorithms, suggesting a new source of theoretical inspiration for creativity researchers.