New Framework Models Interpretive Perspectives in Computational Creativity

A new paper on arXiv (2607.28644) proposes a computational approach to modeling interpretive perspectives in computational creativity, challenging the traditional view of creativity as an objective property of artifacts. The study introduces a four-world framework that organizes a twelve-trait creativity model across four conceptual domains. To operationalize this, the researchers created three evaluative personas—formalist, social-historical, and iconographic—to simulate different critical viewpoints. Using 1,069 artworks from the SemArt dataset, the analysis generated 38,484 persona-based evaluations, revealing systematic divergence in creativity assessments across perspectives. For instance, traits such as 'Social' showed significant variation depending on the evaluative lens. The findings suggest that creativity evaluation is inherently perspective-dependent, and the proposed framework offers a way to account for this variability in computational systems. The paper is categorized as a cross-announcement on arXiv, indicating it may have been submitted to multiple venues. This research has implications for the field of computational creativity, particularly in the development of AI systems that can appreciate or generate art with an understanding of diverse interpretive frameworks.