An Observer-oriented Theory of Creativity and Aesthetic Measure

As a step toward computational modeling of cognitive processes and interpretation of creativity, we present an extended theory of computational creative systems that aims to better define the evaluation of creative systems and processes. Previous research has focused largely on the {\em generation} aspect of creative systems. This paper extends this formalism by modeling the {\em judgment} of creativity through computational models of {\em observer} systems. As a concrete illustration of its applicability, we show how this theory can support the interpretation of MacGyver problems---problems defined in the cognitive systems research community as classical planning problems designed to elucidate the cognitive process of creative problem-solving. Finally, we demonstrate how this theory provides an interpretation to previous empirical work in approximation of an aesthetic measure in artistic domains for individual and group preferences over photographs and gestural performances. Our overall contributions are: (a) an initial formal model of creativity that incorporates a generative process and a given observer (b) application of the theory to the interpretation of results from both symbolic (classical planning) and statistical (preference learning) approaches.