Creative Integration: A Decidable Criterion of Creativity

A new criterion for creative integration proposes that a genuine synthesis between conflicting ideas is one that strictly reduces the description length of the combined concepts, with the reduction located in the conflict itself [1]. The criterion, detailed in a paper posted to arXiv, defines creative integration (CI) as the resolution of a real conflict between A and B if and only if, under a fixed description language, the description length strictly shrinks, yielding a compression ratio C = L_pre/L_post greater than 1 [2]. The work builds on a lineage that treats creativity and intelligence as compression, but the authors state their contribution is not the compression metaphor itself, but its decidability, discrimination, and a supporting corpus [2]. To make the judgment operational, the criterion passes through four binary, conjunctive gates [1]. The authors also construct a taxonomy of pseudo-integration that names and rejects look-alike solutions that might superficially resemble creative synthesis but fail the compression test [2]. The framework is backed by a curated, multi-domain corpus [1]. Validation departs from standard inter-rater agreement methods. Instead, the criterion is subjected to four falsifiable tests: an independent computational check, discrimination against hard negatives, out-of-sample prediction, and description-language robustness [2]. The paper reports that all four tests pass with margin [2]. The authors position novelty and value as downstream symptoms of creative compression, rather than its defining features [2]. This stance intersects with ongoing debates in computational creativity. Prior work has formalized novelty and transformational creativity, proving that novelty is neither necessary nor sufficient for transformational creativity in general, though it becomes a requirement for certain classes of learning agents [5]. Other research has introduced benchmarks like the Creative Invention Benchmark (CrIB), a 2000-problem set spanning five domains designed to evaluate combinational p-creativity—the creativity involved when an individual combines existing knowledge to achieve a novel solution [3]. More recent studies have investigated combinational creativity in vision-language models through frameworks such as Identification-Explanation-Implication (IEI), which decomposes creative synthesis into three hierarchical levels [4]. These efforts share a common challenge: defining creativity in operational terms that can be tested, measured, or falsified. The new CI criterion claims only the sign of C-1 and explicitly judges rather than generates creative output [2]. Whether all creativity is constituted by conflict compression is stated as an explicit conjecture, not a proven claim [2].