The companion essay on ethotechnical design states the positive stance: spend institutional capacity before consuming human adaptive capacity, and keep the arrangement correctable by the people who bear the cost when it fails. This essay draws out the evaluative consequence. If the discipline is right that the allocation of adaptation is politically consequential, then evaluating an institution by its outcomes alone is structurally incomplete, because the same outcome can be produced through radically different distributions of human burden, authority, and adaptation.
The central objection
An outcome tells you what was achieved. It does not tell you what had to happen to people to achieve it. Two hospitals can produce identical recovery rates. Two schools can produce identical test scores. Two organizations can produce identical output. One achieves it through robust systems; the other through exhausted workers, family coordination, unofficial workarounds, and heroic discretion. The outcome measure sees the destination but not the adaptation path.
An outcome tells you what was achieved, not what had to happen to people to achieve it.
That is the structural gap. Outcomes are not worthless; they are incomplete, and the incompleteness is systematic rather than incidental, because the omitted term is exactly the term the discipline exists to make visible.
What a pure outcome orientation rewards
The incompleteness produces failures that compound.
Outcomes can hide compensatory labor. Identical measured results conceal radically different costs, and the measure cannot distinguish the robust institution from the one that runs on its people.
Outcomes can reward cost displacement. A benefits agency that cuts administrative cost by making applicants spend substantially more time documenting eligibility has the same accuracy and a better internal efficiency number. Labor was transferred, and the outcome orientation reads the transfer as improvement.
Outcomes can conceal distributive asymmetry. An intervention can improve average welfare while concentrating adaptation costs on a small group. Aggregate results stay positive; the concentration is invisible unless the measure includes it.
Outcomes do not capture authority. Two institutions can produce identical outcomes while one lets people challenge, force reconsideration, alter rules, and obtain remedies, and the other merely happens to make good decisions. As long as results hold, an outcomes framework cannot distinguish them. But benevolent incorrigibility is not corrigible authority, because today’s good outcome does not guarantee tomorrow’s, and people have a claim not merely to be governed successfully but to hold standing against the power exercised over them.
Outcomes are vulnerable to human compensation. This is the sharpest failure. A badly designed policy can be absorbed by teachers, nurses, caseworkers, and users; outcomes stay good; evaluators conclude the policy works. The better the compensation, the stronger the apparent evidence for the policy.
bad design + excellent human adaptation → good measured outcome → policy validation
The measurement regime rewards the arrangement for consuming the people who prevent it from failing. That is not only a moral problem; it is an epistemic one, because the evidence the institution reads as success was manufactured by the absorption it cannot see. The same structure is the falsified denominator, moved from accounting into evaluation.
Outcomes erase counterfactual capacity. Two systems can produce the same healthcare result while one requires the patient to spend a hundred hours navigating it and the other five. The endpoint is identical; the ninety-five hours are not. Those hours could have been rest, parenting, friendship, paid work, education, or nothing at all. Outcome measurement undervalues the life that could have been lived instead — which is where uncolonized capacity becomes analytically important.
Outcomes cannot show whether the institution can learn. A system can produce excellent results under current conditions while being brittle: suppressing dissent, depending on heroic workers, lacking effective appeals, ignoring edge cases, unable to revise its categories. Current success can coexist with profound vulnerability. Corrigibility is a second-order property — not merely does the institution currently work but what happens when it stops working — and it cannot be inferred from present outcomes.
Outcomes cannot settle legitimacy. This is the deepest limit. Pervasive surveillance might improve a desired outcome; involuntary intervention might improve average health; extreme employer control might raise productivity. An outcomes orientation can incorporate side effects and rights into a sufficiently complicated objective function, but once it does, outcomes are no longer doing the normative work by themselves — prior judgments have decided which outcomes count, whose interests count, what cannot be traded away, and what authority institutions legitimately possess. Effective does not imply legitimate, which is the non-conversion principle reappearing at the level of evaluation.
What evaluation must include
None of this requires rejecting outcomes. That would be an overcorrection: an institution that is wonderfully participatory, corrigible, and nonburdensome while consistently failing its substantive purpose is not successful. The claim is narrower. Institutional evaluation needs at least three dimensions:
What did it accomplish? + What did accomplishing it require from people? + What power did those people have over those requirements?
Or, compactly: outcome + adaptation burden + corrective power. The second term makes the burden distribution visible rather than absorbed. The third term keeps standing in the measure, so that a system whose corrections are nominal is scored for what it is rather than for what it looks like.
The sharpest test of the whole view is the crumple zone. If someone has to destroy themselves to keep the outcome good, the good outcome is evidence neither that the institution works nor that the arrangement is acceptable.
Never evaluate a system without evaluating what people must become, spend, absorb, and surrender for the system to produce its results.
That is the discipline’s deepest challenge to an outcomes-first orientation, and it is a methodological principle rather than a moral one. It does not ask institutions to abandon results. It asks them to state what the results cost the people who produced them, and to keep the measure open to the people who would have to correct them.
What the laws take from this
The method does not require a reader to accept the critique of outcome orientation to adopt its constraints. It takes four, each already bound elsewhere.
From Law IV it takes that every consequential delegation creates a correction obligation, which is the corrective-power dimension of the measure: a delegation evaluated by outcome alone has hidden the correction capacity it owes. From Law VII it takes that error-bearing parties require standing proportional to exposure, which is the political question the outcome measure omits. From Law VIII it takes that observability without state transition is theater, which is the precise failure of an outcome that was produced by absorption: the number is observable, and nothing about it can change the generating rule. And from Law X it takes that the relevant evaluation sits at the highest layer where harm can emerge, which is where the adaptation burden lands rather than where the organization’s boundary happens to be.
An engineer who rejects the essay can adopt all four on narrower grounds: a measure that cannot see the labor its own results depend on is a measure the institution will misread, and a system whose corrections are nominal is not a system it knows how to stop. The theory says why the omission is political. The method only insists that the three dimensions be measured.