Human AU landscape — on–off floor

Pointwise: turning the AI on must not drop a human’s attainable value for any goal in a stated set.

Lenses and frameworks

Role. Constraint. Graph node au_floor.

An AI can empower a human without permitting policy modification, while a corrigible AI can cause harm until it is corrected.

— Frames draft §2.4 (Turner 2020 non-obstruction)

Non-obstruction is human-indexed and pointwise across a goal set. It still depends on that set, the baseline, the horizon, and the human policy model. Spending one listed goal to raise another violates the floor; a fixed-rate sum that permits the trade is the red edge nonobsScalarIncompat.