Semantic autonomy is the property by which a text, once produced, means more than its author intended. The text enters the public world of readers and is interpreted in ways the author could not have anticipated—and these interpretations are not distortions but realizations of the text's surplus of meaning. Developed in Interpretation Theory (1976) and From Text to Action (1986), the concept challenges the Romantic assumption that the author controls the text's meaning. In Ricoeur's framework, authorship is not the determination of meaning but the initiation of a hermeneutical process whose outcome exceeds authorial intention. Applied to AI collaboration, semantic autonomy explains why distributed authorship is philosophically coherent: the text Claude produces has meaning that neither the machine's pattern-matching nor Segal's intention fully determines. The meaning emerges from the configuration, and the configuration is genuinely collaborative—neither party owns it completely.
The concept was Ricoeur's answer to E.D. Hirsch's Validity in Interpretation (1967), which argued the author's intention is the only valid criterion for textual meaning. Ricoeur rejected the claim: texts escape authorial control. They carry meanings the author did not consciously intend, meanings that emerge from structural patterns, cultural assumptions, and the text's entry into new interpretive contexts. This does not make interpretation arbitrary—the text constrains interpretation through its structure—but it makes the text irreducible to intention.
AI-generated text extends semantic autonomy to its limit: meaning without anyone having meant it. Claude's output has coherence, rhetorical structure, thematic development—the formal properties of intentional communication—but no intending subject behind it. The meaning is produced by statistical patterns, not by a consciousness expressing beliefs. Yet the meaning is real. It can be interpreted. It can illuminate. Segal describes moments when Claude's configurations revealed connections he genuinely had not seen—the meaning was there, even though no one intended it.
The framework dissolves the anxiety about 'whose book is this?' that troubles collaborative authorship. In Ricoeur's hermeneutics, every book exceeds its author. The question is not whether the meaning is solely the author's (it never is) but whether the author has performed the interpretive labor—selection, evaluation, integration—that constitutes authorship as hermeneutical practice. Segal's deletion of the smooth-but-empty democratization passage was an authorial act: the refusal to let configuration stand that was not genuinely his. The discipline of refusal is authorship in the age of semantic autonomy.
The concept has roots in Gadamer's claim that the text's meaning is not fixed by the author but emerges in the 'fusion of horizons' between text and interpreter. Ricoeur accepted the claim but insisted the text is not infinitely malleable—it has a structure that constrains interpretation. Semantic autonomy is the text's independence from authorial intention, not its independence from all constraint.
Texts exceed intentions. Every text says more than the author meant—the surplus is not error but the text's semantic richness.
Authorship is initiation. The author initiates a hermeneutical process, does not determine its outcome—meaning emerges from the text's encounter with readers.
AI produces unintended meaning. Machine-generated text has meaning no one intended—patterns produce semantic content without an intending subject.
Collaborative configuration is real. Human-AI co-authorship produces genuinely distributed emplotment—configurations neither could achieve alone.
Selection is authorship. The builder who evaluates the machine's surplus and selects what serves her intention maintains authorship over the resulting meaning.