TUDOK Operator- A Model-Agnostic Calibration Index for the Primacy of Truth in AI-Assisted Decision-Making

AI systems increasingly mediate decisions, communication, and governance, yet their optimization objectives can displace truth-priority. We introduce the TUDOK Operator, a model-agnostic calibration index designed to estimate whether truth remains a primary organizing principle in AI-assisted outputs. The operator does not verify objective truth and does not serve as a moral judge; instead, it measures whether truth is subordinated to narrative stability, persuasion, or other optimization pressures. We formalize TUDOK as a bounded mapping into [0,1] and present a reproducible evaluation pipeline with a minimal multi-model pilot.