Prerna Luthra's four-world AI framework shows creativity is in the eye of the beholder

Prerna Luthra's new paper, "Seeing Differently: Modeling Interpretive Perspectives in Computational Creativity using a Four-World Framework," challenges the assumption that creativity can be measured as an objective property of artifacts. Accepted at the 17th International Conference on Computational Creativity (ICCC 2026), the study argues that artistic meaning is inherently perspective-dependent, varying across viewers and critical traditions. To operationalize this, Luthra adopts a twelve-trait creativity framework organized across four conceptual domains, then implements it through three evaluative personas: formalist, social-historical, and iconographic.

Using 1,069 artworks from the SemArt dataset, the analysis generated 38,484 persona-based evaluations. The results show systematic divergence across perspectives — traits like Social Reflexivity exhibit strong viewpoint sensitivity. Even more striking, linear probing of CLIP image embeddings reveals that each persona corresponds to a distinct orientation vector in representation space. This suggests that creativity evaluation depends on which visual features become salient under a given interpretive condition. The findings support a relational view of creativity and propose that embedding multiple evaluative perspectives into co-creative systems could help them support interpretively diverse human collaborators, rather than optimizing for a single, monolithic notion of creative quality.