A conceptual space, in Boden's technical use, is the structured field of possibilities that defines a creative domain. Chess has a conceptual space defined by its rules, board geometry, and the infinite but bounded set of legal positions. Classical tonal music has a conceptual space defined by harmonic conventions, instrumental ranges, and formal structures. A programming language has a conceptual space defined by its syntax, semantics, and standard libraries. The concept is essential to Boden's taxonomy because the three modes are defined by their relationship to conceptual spaces: exploration searches within one, combination crosses between them, transformation changes them. Without the concept of a conceptual space, the distinctions collapse.
The formulation is not metaphorical. Boden means conceptual spaces in a precise sense: formally definable (at least in principle) structures that specify what counts as a valid move within the domain. The space can be small and tightly rule-bound (chess) or vast and loosely constrained (poetry), but in each case there is something that distinguishes valid moves from invalid ones, novel explorations from mere noise.
The concept lets Boden specify precisely what AI does well. AI excels at exploration within conceptual spaces whose rules can be formalized — which is why chess engines, protein folding systems, and code completion work so spectacularly. Where the rules resist formalization, or where the space is defined by conventions rather than explicit rules, AI exploration becomes more uneven.
Transformational creativity, in this framework, is the operation of changing a conceptual space — replacing one set of rules with another. This is what makes it categorically distinct from exploration or combination. Non-Euclidean geometry did not emerge from exploration within Euclidean geometry; it emerged from breaking one of its axioms and building a different space. Cubism did not emerge from exploration within Renaissance perspective; it emerged from abandoning the single viewpoint.
The concept also explains why some AI-augmented work feels genuinely transformative while most feels merely efficient. The efficiency gains are within existing conceptual spaces. The transformation — if it comes — will involve human builders recognizing that the spaces themselves need to change, and using AI to help construct the alternatives. The machines cannot yet do this work on their own.
Boden developed the concept of conceptual spaces partly in dialogue with computational models of problem-solving (Newell and Simon's problem spaces) and partly in response to Kuhn's paradigm framework. Her innovation was recognizing that creativity could be analyzed in terms of relationships to these spaces — within, across, or changing.
Formal structures of possibility. Conceptual spaces specify what counts as valid within a domain — the rules, conventions, and boundaries that make exploration intelligible.
Not metaphor. Boden means the concept precisely; conceptual spaces are (at least in principle) formalizable structures with determinate boundaries.
The basis of the taxonomy. The three creativity modes are defined by their relationships to spaces: within, between, changing.
Explains AI's pattern. Machines excel where conceptual spaces are formally tractable, struggle where the space resists formalization.
Transformation is framework-change. What makes transformational creativity categorically different is that it alters the space itself, not merely what happens within it.