Consider a simple instruction given to an MI assistant: “Email my professor about the extension.” Without reliable identity infrastructure, the system has to guess which course, which professor, which section, and whether the requester is even enrolled — and it sometimes guesses wrong.
Why “MI,” Not “AI”
The Contact Initiative deliberately uses Machine Intelligence (MI) instead of Artificial Intelligence (AI) wherever reasonably possible, to describe capability that operates strictly within a human-governed structure — able to act only after a person or institution is verified, and only within limits a human governance body has set and can revoke at any time.
How the Authorization Layer Works
Before an MI assistant can act, it must request a scoped grant limited to the specific task at hand — never standing access to a person’s full academic or communication record. If a request is ambiguous, a human-approval threshold is triggered and the assistant asks for confirmation instead of guessing. Every credential is short-lived, and every action is logged for audit.
This is also the foundation for the Machine Intelligence Assistant (MIA) — a proposed persistent, personalized tutoring and mentoring assistant for students and researchers, operating entirely within these same scoped, revocable, and auditable boundaries.