16.11 What would count as scientific creativity?

Running a supplied simulator is not creative. Fitting its parameters is assimilation. Selecting a known intervention more efficiently is useful active learning. Simulator-grounded creativity begins when the system constructs a reusable object withheld from its declaration: a latent mechanism unifying several failures, a measurement making a distinction operational, a factorization separating conflated effects, an experiment macro exposing a new causal contrast, or an ontology extension that transports across regimes.

The artifact must be more than a private latent vector. It must be nameable, typed, executable, and teachable. Its value lies in the family of consequences it makes available, not merely in reducing one loss.

. Pattern discovery proposes what might be true. Simulator-grounded causal discovery asks what would happen under a controlled change. Scientific creativity begins when failures of those changes force the construction of a better mechanism or a better experimental language.