Trisociation

In The You On AI Field Guide

The researchers assume bisociation is a combinatorial operation—that the creative act consists in bringing concepts together, and that more concepts brought together means more creativity. Three is better than two. The machine, which can hold and combine concepts at scale, is therefore a creativity multiplier. The logic is clean, the experimental design competent, and the conclusion wrong in exactly the way Koestler's framework predicts.

The error is the conflation of combination with bisociation—the central confusion of the AI creativity discourse, now formalized in an academic journal. The researchers operationalized bisociation by reducing it to its combinatorial skeleton, stripping away the features that distinguish it from combination: incompatibility of matrices, emotional register of the collision, evaluative judgment that separates structural identity from surface resemblance.

Their empirical finding—that the machine was 'generally more competent than humans at unifying three independent concepts into coherent ideas'—is almost certainly correct and almost certainly beside the point. Coherence is an associative criterion. The machine produces coherent combinations because coherence is what its training optimizes for. But coherence is not the bisociative criterion. A coherent combination of three concepts is not a trisociation any more than a coherent combination of two concepts is a bisociation. The criterion is collision—the perception that the matrices are genuinely incompatible, that their forced contact reveals a structural identity that neither contained.

The paper is a symptom of a broader pattern: the computational creativity community's tendency to formalize Koestler's concepts by stripping them of the features that make them distinctive. The loss is not accidental. It is a consequence of the computational imperative—the requirement that a concept be operationalizable in algorithmic terms. The features of bisociation that resist computation are precisely the features that make it more than combination: the feeling of the collision, the judgment of quality, the emotional charge. Formalizations that omit these features preserve the skeleton and lose the substance.

Origin

The concept was proposed by researchers publishing in California Management Review in January 2025, explicitly framed as an extension of Koestler's bisociation. The paper's reception illustrates how the AI creativity discourse has absorbed Koestler's vocabulary while misunderstanding his distinctions.

Key Ideas

Combinatorial extension of bisociation. The paper treats creativity as concept-combination, making three better than two.

Formalization loses substance. The features that resist computation (felt collision, quality judgment, emotional charge) are the features that matter most.

Coherence is the wrong criterion. The machine produces coherent combinations easily; coherence is associative, not bisociative.

Case study in the fluency trap. A respected journal publishes a flawed formalization because the error is invisible to associative evaluation.

Diagnostic of the discourse. The paper's publication shows the AI creativity community lacks Koestler's core distinction.