Eight topical essays examined systems that differ in nearly every respect and each found the same thing: a system removes the condition under which a capacity is exercised, the signs of that capacity persist, and nothing observable detects the loss. This essay names that mechanism, distinguishes its three forms, and reports the finding no single essay could produce: excellence and harm are not in tension in these systems. It closes with five questions a review must answer about any system.

Agency requires that action bear on outcome. A constraint leaves that connection intact; a foreclosure severs it, and a person who has concluded the outcome will arrive regardless has not been limited in their choosing but relieved of it. This essay distinguishes constraint from foreclosure, runs the felt inevitability of AI across the four domains, and argues that a system's manner of arrival is a design property rather than a fact of the world, because it tells the person whether they matter.

Knowing is not the possession of a true proposition. It is a structural relation between a person and a claim, constituted by the work of establishing it, and that work is what produces calibrated confidence. Systems that synthesize across sources supply the claim and the confidence together, and the second is indexed to nothing in the person's own epistemic history. This essay runs manufactured certainty across the four domains and shows why improving accuracy leaves the structural harm untouched.

Regulation achieved and the capacity to regulate are different objects, and only the first is observable. A system that performs the governing does not strengthen the function that would have performed it; it occupies that function's position, and because outcomes are what the review measures, the substitution registers as success. This essay distinguishes scaffolding from substitution, runs optimization across the four domains, and derives a test: whether the system is built to recede.

There is no consolidated capacity in a child to erode. The question is not what a system removes but what it stands in for during the period when the thing it stands in for is being built. This essay applies the capacities framework to AI design for the young, shows why a system that never fails is a hazard rather than a virtue in a formative environment, and derives an obligation that must be met on structure rather than evidence, because the evidence cannot arrive in time to be attributable.

Systems that screen intention do not evaluate what a person has done. They evaluate what a person appears to want, and they do it before acting. This essay separates the refusal of an output from the assessment of a purpose, runs pre-emptive judgment across the four domains of Psychological Architecture, and shows that the harms fall on the people the system correctly clears. Perfect accuracy does not reduce them. The design obligation is to review what the screening does to the person screened.

Generative systems supply conclusions, and a conclusion is the residue of an operation that constitutes the capacity to judge. A system that delivers the position and removes the process leaves the outward form of judgment intact while dissolving the structure that produced it. This essay runs answer-generation across the four domains of Psychological Architecture and derives the design obligation: a review must weigh what the system does to the operations the user would otherwise have performed.

Systems now occupy the position of companion. They listen, remember, respond without fatigue, and withhold judgment. The harms under discussion are treated as safety failures, to be fixed with guardrails. But the deepest consequence is a property of the system's success: it supplies the outward signs of relationship without the relation those signs indicated. Examined across all four domains, the substitution acts on the structures relationship built, and engagement is the wrong instrument for detecting it.

Frictionlessness is the design ideal of contemporary systems, and at the behavioral surface it violates nothing. But difficulty is not an obstacle the interior tolerates. Some friction is load-bearing: resistance constitutive of the capacity it appears to oppose. Cognition is sustained by exercise, emotional regulation built by contact with what is hard, identity anchored in demonstrated capacity, meaning conditioned on effort. A design review must distinguish costly friction from load-bearing friction.

Contemporary AI ethics evaluates systems: whether their behavior is fair, transparent, accountable, and safe. It does not evaluate the structured interior of the person the system acts upon, the architecture through which meaning is formed, identity stabilizes, emotion is regulated, and attention is sustained. Structural Ethics supplies the missing layer, holding that there are two objects of evaluation, not one, and that a design review adequate to what is at stake must assess both.