A Formal Framework for Defining and Measuring Creativity in AI Applications

Summary

Computational creativity still lacks a definition that is conceptually grounded, practically useful, and empirically testable across domains. This paper addresses that gap by proposing formal postulates based on psychological research, particularly Margaret Boden's three criteria of creativity: newness, value, and surprise. It then develops a conceptual mathematical definition aligned with those postulates. The authors validate the formula in two applications: evaluating the creativity of individual chess moves and assessing paintings by artists including Picasso and Renoir. The experiments report that the formula captures creative patterns in both chess and art. The results suggest that creativity can be represented by a unified mathematical structure, while the way newness, value, and surprise are measured must be adapted to each application domain.