— Ch. 1 · Defining The Creative Machine —
Computational creativity.
~7 min read · Ch. 1 of 7
In 1956, the Dartmouth Conference listed creativity as a key goal for artificial intelligence. This moment marked the first time researchers explicitly asked if machines could be creative. The field of computational creativity now sits at the intersection of artificial intelligence, cognitive psychology, philosophy, and the arts. It seeks to model human-like creative processes using computer systems. Some goals involve constructing programs capable of human-level creativity. Other aims focus on understanding human creativity itself through an algorithmic perspective. A third strand designs programs that enhance human creativity without being creative themselves. Theoretical work runs parallel to practical implementation. One strand informs the other in a continuous loop. Margaret Boden defined creativity as the ability to produce ideas or artifacts that are new, surprising, and valuable. She focused on two criteria: novelty and value. Newell, Shaw, and Simon added usefulness to their four-part definition. They required answers to be novel and useful while demanding rejection of previously accepted ideas. Mihaly Csikszentmihalyi argued that creativity must exist within a social context. His DIFI framework describes how individuals produce works assessed by society. The field distinguishes between P-creativity and H-creativity. P-creativity refers to psychological novelty for the agent producing it. H-creativity denotes historical novelty recognized by society at large. Exploratory creativity arises from searching within established conceptual spaces. Transformational creativity involves deliberately transcending those same constraints.
Early Symbolic Models And Systems
During the late 1800s, methods for composing music combinatorially involved prominent figures like Mozart, Bach, Haydn, and Kiernberger. Simple mechanical models were built as early as 1934 to explore mathematical problem solving. Meehan's TALE-SPIN system appeared in 1977 to generate narratives through simulation of character goals. Dehn's AUTHOR system arrived in 1981 to approach generation by simulating an author's process. Harold Cohen's AARON began producing art in 1973 through composition and adaptation of figures based on symbolic rules. BACON rediscovered natural laws like Boyle's Law and Kepler's law through hypothesis testing. Turner's MINSTREL system introduced TRAMs in 1993 to simulate creative re-use of prior material. Pérez y Pérez's MEXICA model followed in 1999 to model the creative writing process using cycles of engagement and reflection. The JAPE system emerged in 1994 to generate pun-based riddles using Prolog and WordNet. David Cope's EMI software trained itself on styles of artists like Bach, Beethoven, or Chopin. These systems relied on explicit formulation of prescriptions by developers combined with randomness. They used rule-based approaches to generate creative artifacts. Professional interest in the creative aspect of computation was commonly addressed in early discussions of artificial intelligence.