AI Interrogation

Health technology assessment (HTA) has spent decades constructing a numerical architecture that appears authoritative but lacks the most basic requirements of scientific measurement. The result has been a profession sustained by conventions, utilities, QALYs, DALYs, and reference-case simulation models that cannot be validated, replicated, or defended under the axioms of representational measurement theory. The AI Interrogations project was created to expose this foundational failure. By systematically questioning large language models trained on the global HTA literature, we reveal what the field has internalized, what it has forgotten, and crucially, what it has never understood.