Deliberate Incoherent Prompting
A deliberately incoherent, heterogeneous prompt assembled by the user so that the AI system produces unexpected or creative combinations from the disparate elements. Distinguished from an ordinary prompt by its intentional internal inconsistency.
Operational Definition
EVENT. One occurrence would be coded when a user prompt deliberately juxtaposes three or more semantically unrelated elements (topics, registers, or domains) in a single request, as judged by coders, with the apparent aim of eliciting novel combination. The composite prompt is the unit counted.
Measurement Schema
Proposed measurement protocol (not yet empirically validated): Prompt-level content coding: raters count distinct unrelated elements per prompt and flag intentional juxtaposition; agreement would be assessed via Cohen's kappa. Complementary objective index: lexical-semantic dispersion of the prompt (mean pairwise embedding distance of its content segments). Could be reported as deliberate-mix prompts per session and mean dispersion.