T(A)
Telos
The end for which the agent acts. Machines optimize objectives; only humans hold ends.
Theoretical Foundation
Agency Anchoring: one apparatus, every domain where AI acts.
Agency Anchoring (AA) is ERTC's answer to the central question of the AI era: when an artificial agent can perform an expert's operations, what — exactly — must the human still do? AA formalizes agency as a small set of primitives, then derives, domain by domain, which operations may be outsourced and which must stay anchored to a human who can answer for them.
R(A)
Answerability
Who stands behind the act and bears its consequences. Answerability cannot be delegated to what cannot be held to account.
Ω(A)
Operation set
Everything the agent can do — the full repertoire of operations available in a domain of practice.
Ωₕ(A)
Anchoring set
The non-empty subset of operations that must remain with the human. Identifying Ωₕ(A) is the analytical work of each domain instantiation.
HAA
Education
Human Agency Anchoring in learning. AI may tutor, scaffold, and assess — but the learner's questioning, sense-making, and the educator's judgment are non-delegable. HAA is the design law behind every ERTC platform.
SAA
Scholarship
Scholarly Agency Anchoring. If an AI system satisfies telos-bearing and answerability conditions within a supervised anchoring structure, what authorship roles can it legitimately hold? SAA gives the field a principled test instead of a taboo.
MAA
Medicine
Diagnosis support may be outsourced; the clinical encounter, informed consent, and the duty to the patient stay anchored.
JAA
Law
Research and drafting may be delegated; judgment, proportionality, and accountability before the court cannot.
PAA
Manufacturing
Process control scales through automation; responsibility for safety cases and quality ends remains human-held.
DAA
Autonomous driving
Perception and control are machine operations; the definition of acceptable risk is a human anchoring operation.
制可外包，知须锚定 Execution can be outsourced; knowing must be anchored.