Rhetorical-Register Discrimination Ability
An individual user's graded ability to identify the rhetorical register of an AI output, for example whether it is serious, ironic, neutral, or exaggerated. Distinguished from comprehension of content by its focus on tone, and it varies across users.
Operational Definition
STATE. The rated quantity is a user's tone-classification accuracy: across a fixed set of AI outputs pre-labelled for rhetorical register by expert consensus, the proportion the user labels correctly. Higher proportion indicates greater ability. Measured per user, not per interaction.
Measurement Schema
Proposed measurement protocol (not yet empirically validated): Forced-choice tone-classification test: each user labels a calibrated item set against an expert-consensus key; ability scored as percent correct (or d-prime for irony vs literal). Item-set reliability reported via Cronbach's alpha; expert key reliability via Fleiss' kappa across labellers.