Specification vs. Interpretation in AI Collaboration

In Goodman's analysis of allographic arts, notation maintains a precise boundary between what the composer specifies (identity-determining features like pitch and rhythm) and what the performer interprets (features the notation leaves open like timbre and micro-timing). The boundary is formal—determined by which features the notation can specify with differentiation and unambiguity—and it is what allows musical works to survive across performances with their identities intact. In AI collaboration mediated by natural language, the boundary dissolves. Natural language lacks the formal properties Goodman required of notation—it is not syntactically differentiated (words belong to multiple categories), not semantically unambiguous (sentences admit multiple interpretations), and not precise enough to distinguish features the human is specifying from features the human is leaving open to the machine's judgment. The result is that the machine's rendering can introduce features the human did not intend—new claims, structural choices, examples, emphases—and the introduction is indistinguishable from the filling-of-interpretive-space that a performer legitimately does. The human specifies an intention; the machine returns a rendering that has