Combinational creativity produces novelty by making connections between ideas that are familiar in isolation but have not previously been combined. The poet who connects grief to geology, the scientist who borrows a mechanism from evolutionary biology to explain technology adoption curves, the engineer who recognizes that laparoscopic surgery and AI-augmented coding share a structural pattern — each performs combinational creativity. The mode is distinct from exploration because it crosses conceptual spaces rather than searching within one. AI is a powerful engine for combinational work because its training corpus spans the documented range of human knowledge, enabling connections no individual mind could make because no individual has read everything. But the quality of a combination — whether it produces genuine insight or merely surface cleverness — requires evaluative judgment the machine cannot supply.
The classic human example Boden cites is Einstein's connection between the behavior of clocks in different reference frames and the behavior of light — drawing an analogy across domains that no one had previously linked. Dylan's 'Like a Rolling Stone' emerged from a similar process: combining traditions — folk narrative, blues compression, Beat poetry — that had existed separately. Edo Segal's punctuated equilibrium connection that emerged in conversation with Claude — linking evolutionary biology to technology adoption — is a textbook case.
The mode has an asymmetric quality: generating candidate combinations is computationally tractable (especially at AI scale); evaluating which combinations are genuinely illuminating requires judgment that draws on taste, domain depth, and cultural context. This asymmetry is the structural basis for the human-AI partnership that The Orange Pill describes. The machine produces the connections; the human recognizes which deserve to exist.
Combinational creativity is what most human-AI collaboration actually produces in the current moment. The language model surfaces a concept from an unexpected domain; the builder recognizes it as illuminating or dismisses it as empty analogy. Neither could produce the final insight alone — the machine lacks the evaluative capacity, the human lacks the range. The quality of the output depends on both partners, and especially on the human's capacity to tell genuine connections from plausible-sounding ones.
The risks of combinational creativity at AI scale are significant. The machine generates many superficially plausible connections that do not hold up under scrutiny — what the Deleuze fabrication in The Orange Pill illustrates. Without the human's disciplined evaluation, combinational output degrades into a kind of sophisticated noise: prose that sounds like insight without actually being insight.
Boden developed the distinction between exploratory and combinational creativity partly in response to Arthur Koestler's bisociation framework in The Act of Creation (1964). Koestler had proposed that creativity involved the collision of previously independent matrices of thought. Boden recognized that this mechanism — while real — described only one mode of creative production, not all of it.
Cross-domain connection. The defining operation is linking ideas across conceptual spaces rather than searching within one.
AI's structural advantage. Training corpus span makes combinational work accessible at ranges no individual human can match.
Generation is cheap, evaluation is not. The machine produces thousands of candidate combinations; the human must supply the judgment that separates illuminating connections from superficial ones.
Quality depends on partnership. The best combinational work in the AI age emerges from sustained collaboration between the machine's range and the human's evaluative depth.
The risk of empty analogy. Without disciplined evaluation, AI-assisted combinational work degrades into plausible-sounding prose that collapses under examination — as with the Deleuze fabrication.
A persistent question: at what point does sophisticated combination shade into transformation? Boden's position: even radical combinations remain combinations if they operate within the existing rules of the combined domains. Transformation requires changing those rules, not merely linking them. Critics argue the boundary is fuzzy in practice, though Boden maintains it is categorically clear in principle.