The Ehstand Test: Can a Machine Faithfully Continue One Person's Thinking? Canonical Definition, Score Card, and Protocol (v1.0)

Abstract

Canonical founding document of the Ehstand test (v1.0, bilingual English/German). The Turing test asks whether a machine appears generically human. The Ehstand test asks a different question: can a machine faithfully continue the thinking of one specific person — their concepts, held positions, reasoning moves, transfers to new cases, updates under new evidence, refusals, and dialogue continuations? As digital delegates, knowledge succession and "this agent thinks like X" claims spread, such claims need a test. This working paper states the canonical five-sentence definition, the EHSTAND score card (seven dimensions), the public scale Ehstand 0–5, the measurement protocol (sealed personal archive, commit–reveal, blind rating against baselines, passing criteria), the task families with weights, limits and open questions, and the relation to prior work. Name use: the term is openly citable as "Ehstand test (Ehstand, 2026)"; independent test runs are encouraged. "Certified Ehstand" remains reserved for protocol-conformant assessments. Licence note: CC BY-ND 4.0 covers the canonical text; trademark and name rights remain unaffected; the research programme's internal working procedures are not part of this publication.