Objections and Responses

This is human-in-the-loop with a philosophy attached. A human reviews the model's output. That framing already exists. Nothing here is new.

The objection inverts the actual claim. Human-in-the-loop places a human beside a system that performs the cognition — the model does the work, the human checks it, the human is a safety valve on a fundamentally autonomous process. HI-Centric-AI holds the opposite. The named human performs the part of the cognition current artificial intelligence cannot: the simultaneous, multi-level, experiential integration of judgment within a domain. The system does not do the thinking and submit it for review; it extends the reach of a judgment it is structurally incapable of producing. The human is not supervising the intelligence. The human is the intelligence, and the system is its instrument.

This is Engelbart relabeled. Augmenting Human Intellect said all of this in 1962. Naming it again is not a contribution.

The discipline claims no novelty in the augmentation insight and attributes it to Engelbart directly. The contribution is its articulation for a context Engelbart never addressed: regulated professional practice, where named human authority is not merely useful but constitutive — mandated by license, profession, or institutional bond, and legally inseparable from the work. Augmentation extends generic human capability. Amplification centers a specific named expert because that expert supplies the integrated judgment the system otherwise lacks, and because, in these domains, that expert cannot be removed without dissolving the accountability the practice is built on.

The lineage ends in 1969. If this tradition were alive, it would not need reviving. The field moved on to autonomy for a reason.

The tradition did not end; it was eclipsed. The augmentation and symbiosis programs continued through human-computer interaction, intelligence-augmentation research, decision-support systems, and the human-centered artificial intelligence work of the present era. The discipline does not revive a dead field. It consolidates a living but scattered tradition into a named, practiced discipline for a specific application: high-stakes regulated domains where autonomy is not the appropriate goal. The autonomy program advanced for good reasons and serves the problems it suits. It does not suit every problem. The discipline exists for the class it leaves unserved.

This is unfalsifiable. It is philosophy dressed as a discipline. Nothing here could be shown to be wrong.

A discipline that cannot name the conditions under which it would revise itself is a creed. HI-Centric-AI names them. A system fails to qualify where authority is anonymous rather than named, where the knowledge substrate is unbounded rather than scoped and inspectable, or where it depreciates rather than compounds under expert use. The deeper premise is testable too: the discipline centers the human because current artificial intelligence works from imposed knowledge, not lived experience, and cannot integrate multi-level judgment as an embodied expert does. Should an artificial system come to hold its own vessel and its own experience — to integrate judgment as an organic intelligence does rather than from a corpus it was handed — the central premise would require honest revision. The discipline does not claim this is near. It does not claim it impossible. It holds the question open.