Complete bounded explanation
What this power means.
Capability to develop or express conceptual blending in a declared context without inheriting broader claims.
What the current evidence supports
Conceptual combinations can be generated and evaluated on declared tasks; general trainability and real invention transfer remain unproven.
How to observe or measure it without overclaiming
Independent novelty, coherence, feasibility, provenance and user-value ratings on a constrained blend task. Judge scores reflect a declared rubric and audience, not an objective universal essence of creativity.
Combining concepts automatically creates a breakthrough
MetricIndependent novelty, coherence, feasibility, provenance and user-value ratings on a constrained blend task
BoundaryIndependent novelty, coherence, feasibility, provenance and user-value ratings on a constrained blend task. Judge scores reflect a declared rubric and audience, not an objective universal essence of creativity.
Negative and limiting findings
- Analogy prompts did not reliably trigger transfer, and the Fusion Innovation Test is measurement evidence rather than acquisition evidence.
- No evidence supports perfect, limitless or universal performance.
Claims this evidence cannot support
- Mixing any two ideas creates innovation
- AI-generated blends are automatically the user's authorship
- Novelty alone establishes value