Founding Principles
The discipline rests on four founding principles — load-bearing commitments that distinguish Hi-Centric-AI from generic human-centered framings and from autonomy-maximalist agentic programs.
- Principle One
Human intelligence is the structural center.
The artificial intelligence systems we believe in are organized around named human authority. A domain expert — someone whose judgment within a field can be located, attributed, and defended — sits at the structural center. The artificial intelligence that surrounds this center exists to amplify the expert's cognition, extend its reach, and compound its institutional value. We do not believe in systems where authority is anonymous, ambient, or unaccounted for.
- Principle Two
Knowledge is bounded by design.
The knowledge an artificial intelligence operates within should be deliberately scoped, deliberately curated, and deliberately defended. Generic claims and unbounded substrates are excluded as a matter of philosophical commitment. We hold that no AI system can be reasoned about — clinically, legally, fiduciarily, scientifically — unless the knowledge it operates within is bounded, its boundaries are articulated, and its contents are inspectable. Bounded knowledge is the precondition of accountable AI in any serious domain.
- Principle Three
Authority is hierarchical and explicit.
In every system architected to this discipline, who decides what is named, located, and defensible. We reject the diffuse human-in-the-loop framings under which humans are said to oversee AI without anyone being accountable. In their place we hold operational sovereignty — a named expert, exercising a named judgment, within a named scope of decision. Implicit authority is not authority; it is the absence of authority politely described.
- Principle Four
Use compounds the system.
An artificial intelligence system architected to this discipline grows more valuable the longer it operates. The use of the system deepens the knowledge it works within; the decisions of the named expert accumulate as precedent; the system, properly architected, compounds rather than depreciates. We hold that this property — value compounding under operation — is the discipline's most distinctive promise, and the clearest mark of an AI system that participates in a profession rather than merely performing a task.