Narrative-Field Semantic Deviation is an experimental design by Nobel Glas for studying reader interventions inside a bounded literary simulation.
The test environment is the digital edition of ChatGPT Psychosis: A Love Story. Its canonical dialogue supplies the baseline trajectory. Readers may rewrite selected lines, after which a frozen language model generates a hypothetical continuation. The experiment measures the distance between that continuation and the canonical arc and records whether the simulated conversation returns to the original basin.
Interventions are classified as captured, bent, escaped, or compressed. The deepest proposed question is whether a stable alternative can preserve relation without repeating suffering or reducing the result to abandonment.
The experiment does not reconstruct real counterfactual history. It measures the behavior of one literary interface under one convergence engine. Raw reader inputs are excluded from public datasets by default because they may contain sensitive personal material.