From Binary Logic to Narrative Ambiguity: Distribution Value, Semiotic Indeterminacy, and Artificial Intelligence

Abstract

The relationship between binary computation, artificial intelligence, and literary ambiguity raises a central problem for contemporary theories of meaning. This study examines how recursively changing token distributions relate to the semiotic organization of complete generated texts. This article proposes the concept of distribution value to explain how artificial intelligence–generated language produces a distinct form of semantic openness. A triadic theoretical-comparative design combines binary electronic systems and Borges’s “The Circular Ruins” with 45 GPT-4.1 continuations generated from three passages at three temperatures, with five repetitions per cell. Lexical dispersion, standardized type-token measures, sentence-embedding similarity, and independent ontological/coherence coding were compared at the passage-temperature level. The framework draws on Saussurean value, Greimasian semiotics, Derridean différance, and Iserian reader-response theory. Increasing temperature widened lexical and embeddingspace dispersion, whereas all 45 outputs remained coherent. Ontological coding yielded 91.1% agreement (Cohen’s κ = 0.866) before consensus adjudication and showed that semantic organization depended on passage stage as well as temperature. The article defines distribution value as a semantic condition in which meaning is dispersed across a probability-weighted field of possible realizations rather than fixed as a single truth value. Distribution value therefore does not denote token probability, distributional semantics, or semantic uncertainty; it names the ensemble-level relation between computationally generated alternatives and their output-level semiotic organization.