Who pays for this work
The essays are free to read. I offer clinical AI safety reviews to health plans, healthcare AI companies and investors, listed on kanav.net. I also offer advisory and fractional product work to healthtech companies. I’m building NextConsensus, a commercial product that tracks when medical evidence moves ahead of clinical guidelines. I previously built healthcare products at Doximity, Transcarent and Epic.
The open standards at ethotechnics.org are free under CC BY-SA 4.0; the reviews would be the paid work. Health plans are among the buyers the reviews are offered to, and in several essays I criticize prior authorization and the insurers who run it. If one reads as going easy on a would-be buyer, or as selling a service, tell me.
About Kanav Jain
A nurse keeps a ward running by fixing failures the hospital's records never show. A patient knows exactly what's wrong and still has no way to change it. I study what that sustaining work does to people's skills, obligations, alternatives, and authority. I ask how institutions judge their own performance, whose labor those records leave out, and who can force a correction. Adaptation can lock a bad setup in, or cooperation can build shared authority; I study what splits those paths and what gives an institution a claim on people's future work. That question owes a lot to republican non-domination and social reproduction theory; the lineage page records those debts with the rest.
The governing question is who gets to spend whose life. Human capacity is finite and institutions make mistakes, so power decides whose limits an arrangement accommodates and whose adaptability it extracts as a free subsidy. Production is a means; the aim is the greatest capacity for living.
Progress that expands people’s ability to shape the future
I think the people a system serves should have authority they often don’t have now: a route to challenge a decision, a fallback when the system is wrong, and a say in who can change the rule.
So I’d judge a technology or an institution by whether it leaves people better able to do those things, as well as by how well it serves them now. That doesn’t mean people should get whatever they want, and some decisions shouldn’t be put to a vote.
I don’t know how much of an arrangement should stay open to change. What I don’t accept is that an institution’s dependence on someone’s past work gives it a claim on their future work.
Who can change a system that works?
A worker covers a staffing gap and the service keeps running. The next schedule can assume the same extra effort. I ask who does that work and what it does to their power to change the arrangement.
The goal is institutions strong enough to run shared life that cannot use people up to keep themselves running.
Central hypothesis
What the program studies
Staff work around a broken rule, and the institution stops feeling pressure to fix it. Meanwhile, those workarounds burn the time and energy people would need to change the setup. That loop is the foreclosure this program studies. I trace it through workarounds, what the institution’s records count, and which decisions people can actually change.
The test runs over time: who gains skills, who loses alternatives, and who can set new terms? Good support can widen those options. A care service gives someone more freedom when it delivers reliable help and gives them a real say in how it runs.
Knowing a problem exists does not change it. Who enters the record depends on who can afford the unpaid work of reporting. A practice can persist after its justification weakens, and better data rarely hands anyone the authority to act.
Six distinctions recur: improvement vs. progress, assistance vs. authority, transparency vs. correction, innovation vs. dependency, formal vs. effective accountability, and shallow vs. recursive corrigibility.
Each mechanism has published evidence behind it; the whole process is untested.
Every improvement faces one design test: does it leave the people it serves with more authority over the institution, including the power to replace it or need it less? Put more formally, under what conditions can institutions grow stronger while the capacity to initiate, fund, evaluate, and carry out alternatives stays with the people they affect?
We should not have to preserve the conditions that shaped us
My predictions track institutions getting people wrong: a claim denied, a record nobody can fix, a rule that repeats the same failure. Behind them sits one puzzle. Institutions are growing faster at reshaping the world than the people affected are at correcting them.
Three concepts organize the program. The politics of necessity explains how contingent arrangements come to feel inevitable. Adaptive foreclosure explains how people's successful workarounds make those arrangements harder to replace. Effective revisability is the practical power to change the terms governing your life.
The first two are causal hypotheses about how arrangements persist. The third is a criterion, not the whole standard: I argue for justified authority over what continues and what changes, and revisability is how the people affected test it.
The politics of necessity
People organize life around a contingent setup. They adapt, invest effort and identity, and build dependencies. Alternatives grow costlier to pursue, until the setup looks inevitable and its own fallout becomes the argument for keeping it.
The sharpest version is manufactured validation. An institution doesn’t just call a choice destiny; it makes the world confirm its assumptions. A platform defines engagement, shapes user behavior to match, and cites the metric as proof its design works. A hospital raises productivity targets until staff absorb the load, then treats their survival as proof the targets were fair. Producing confirming evidence crowds out independent proof against the institution’s assumptions.
Some instruments survive being turned into targets, and a measure can hide what it leaves out before anyone games it. Inevitability has to manufacture its own proof.
How can an institution tell the difference between discovering it was right and gaining enough power to make itself right?
Adaptive foreclosure
People adapt. Nurses get good at covering understaffed wards. Workers learn to route around broken software. Communities build informal networks when public services fall short. That competence is real—and it hides the underlying failure from the people who could fix it. Barbara Levitt and James March called one version a competency trap: the better an organization gets at an inferior procedure, the more it pays to keep using it.
I call the result institutional insulation. The better people get at living under a constraint, the easier it is for the institution to keep it. Survivability measures what refusing that arrangement costs. Developing a skill under pressure should not obligate you to keep supplying it forever; a cover offered once can become a plan that can't run without it.
There is a counterweight. A person shouldn't have to earn their place by being useful, and a refusal doesn't have to prove the institution wrong. Whether a change improves the system is one question; whether you must keep carrying it is another. Rank spares some people from adapting at all, a person has a right to get tired, and a promotion can make criticism more expensive, from the tenure settlement to the Freedmen's Bureau.
The evidence follows the same loop. Selection, socialization, and dependence decide who stays and which disagreements feel safe. The people who call an arrangement legitimate are the ones it kept.
Foreclosure runs on two tracks. First, adaptation: people absorb the cost until the setup looks self-sustaining. Second, the record: institutions keep thorough files on the moment you agreed, because that file enforces the contract, while keeping no file on whether the agreement was justified. That missing file shapes the institution's account of itself.
Effective revisability
Counting options on a menu does not measure freedom. Workers can switch employers without shaping how work is organized. Staff can request an accommodation without contesting the norm, or hold a formal right to refuse that costs too much to use. Freedom requires the practical power to contest and revise the conditions behind the options—what constitutional friction and reversibility demand.
The test is whether you can still change an arrangement after you depend on it. Picking from offered options is first-order freedom. Helping decide which options exist is second-order freedom. An arrangement can widen the first while narrowing the second, and this test tracks the second.
Revisability is not the same as restitution. A system can roll back a record, correct a decision, repair the fallout, restore losses, or change the rule that caused the failure. Many succeed at the first two and stop. Legitimate automation requires operators who can carry responsibility for harms that cannot be undone.
Sometimes the protective move is foreclosed in practice; sometimes the option exists and using it costs too much. One measure of freedom is the power to revise the arrangements you have adapted to.
The legitimate obligation
Contestability has a cost. A promise you can drop whenever it turns inconvenient is not a promise, and a decision that never settles anything cannot govern. The hard question is how people can rely on one another without locking someone’s freedom away—the line conditional authority draws.
The sharpest objection to this program is simple: how much revisability is enough? There is no single number. The real question underneath is who holds authority over which revisions, through what procedures, and with what duties to the people affected. Any serious answer has to separate legitimate stability from domination by weighing expectations, compensation, adjustment costs, and collective authority.
Stability counts as evidence for an arrangement under three conditions: people can leave without ruin, benefits reach the people carrying the costs, and the terms stay open to revision by those bound by them. Most durable arrangements meet those conditions only partly. Persistence proves neither legitimacy nor illegitimacy on its own; the conditions decide.
You can take pride in surviving hardship without arguing that the hardship should stay. The value of what you endured does not require preserving the conditions that caused it.
The test
The usual fear about capable AI is losing technical control of the machines. The likelier failure is keeping technical control while losing political and practical control over the institutions deploying them.
The test is whether governed people can change the conditions that make an institution look necessary. New capabilities are one half of progress. The other half is who holds the effective freedom to pursue and revise them: the people affected, incumbent organizations, or public authorities.
Two failures get blurred here. In the first, an objection never reaches the decision, so the rule looks uncontested only because nobody could contest it. In the second, a well-founded objection arrives and the rule survives anyway. One is a broken channel; the other is an unaccountable decision. An institution can listen well and still refuse to act—and what the first failure keeps off the record is often the proof the second one happened.
A third failure sits outside the institution. Everyone agrees the change is justified, yet the people affected still can’t make it because transition costs land on whoever depended on the old setup. An institution can pass every test on this page and still be unable to end if the surrounding economy can’t absorb the exit.
The hardest open question is how institutions preserve that freedom without making durable cooperation impossible. That is where critique has to become institutional design.
What would count against this
A framework that can explain every case has stopped being a framework.
- Authority builds up through the interfaces people have to use (forms, screens, scripts), not only through formal rules. It would count against this if two institutions with nearly identical interfaces held very different power over people, and the law alone explained the difference.
- Foreclosure (people losing the means to change a system) and entrenchment (the system becoming hard to remove) are separate failures. It would count against this if, across many cases, they always rose and fell together, because then they would be one failure under two names.
- Institutions rarely keep a record of why a rule was justified. It would count against this if many kept such records without being required to.
- Adaptation makes an institution durable. It would count against this if institutions stayed durable while the people under them gained the capacity to replace them.
- Correction depends less on whether an appeal exists than on whether fixing a case changes the rule. It would count against this if systems with weak appeals turned out to change their rules often.
The adaptation test is the one to watch. If every case of institutional success gets read as incipient foreclosure, the framework has reproduced the failure it studies, and the correction should appear here rather than in a footnote.
Read through the argument
Each essay below takes the question a step further, from the work people supply to the terms of continued cooperation.
What are people doing to make this work?
Start with the benefits and the work that supplies them. A nurse's repair can protect a patient while leaving the rule that required it unchanged. Ask who may refuse that work and who must keep absorbing it.
What does their success appear to prove?
A completed task shows that someone got it done. It doesn't, by itself, establish adequate staffing, reasonable demands, or consent to the arrangement. The contributions and their costs belong in the assessment of performance.
How did their contribution become something they owe?
Being capable, appreciated, or relied on can make a contribution difficult to stop. That history doesn't by itself settle its scope or duration. Promises and responsibilities to others need examination alongside the expectations that accumulated.
Read Being Good at It Doesn't Mean You Have to Keep Doing It
What happens when they challenge the terms?
Support and professional success can expand someone's opportunities while raising the cost of disagreement. Ask what security and alternatives remain when their judgment conflicts with the institution's interests.
What would make continued cooperation justifiable?
Some duties sustain valuable shared life. Ask whether the work is bounded, shared, replaceable, and paid where appropriate, and whether those doing it can revise the terms. A remedy needs the same scrutiny as the arrangement it changes.
Four questions this research asks
The homepage asks three questions of an institution: Distribution, Correction and Jurisdiction. The four lenses below include those three and add reproduction, how a setup keeps itself in place. The homepage does not ask that one, and its three tests still stand. A pattern can answer more than one lens.
Where does the cost go?
Distribution follows a cost after the institution stops counting it. A faster process may eliminate work or move it onto someone else. I trace the total, including the risk people carry outside the accounts, and how it changes over time. Burden answers this most directly; infrastructural power and the defect story show the other ways a cost gets placed and blamed.
Where does authority stop being legitimate?
Authority asks what an institution was entitled to decide, and what ends that entitlement. The cost of saying no can change as people depend more on a system, so I ask whether they can refuse or renegotiate its demands while keeping access to care and income. Conditional authority answers this most directly; survivability and vibe-based legitimacy show how refusal gets priced and how a likable surface keeps power hard to argue with.
Read the conditional authority pattern · Diagram: Being Right Doesn’t Make It Yours to Decide
How does the institution preserve itself?
Reproduction asks how an institution keeps going once its costs are known. A completed task can hide the repair that made it possible, so I ask whether the institution records that work, whether it has the means to remove its cause, and what happens when it knows the cost but leaves the rules intact. Care theater answers this most directly; delay and burden show the other ways a system survives without changing.
Read the care theater pattern · Diagram: When Covering Becomes the Plan
Who can change the rules?
Transformation asks who can revise the institution, not only correct one decision. Correcting a file may leave the rule that produced the problem unchanged, so I ask what the people affected can change, who controls the resources needed to do it, and whether their authority grows or shrinks as the institution develops. Constitutional friction answers this most directly; reversibility and conditional authority show what stopping points and renewal terms make change possible.
Read the constitutional friction pattern · Diagram: Fixing the Case, the Rule, or Who Decides
How this work relates to other research
The work draws on systems engineering and political philosophy, including traditions recorded in the archive's lineage. It uses engineering questions about maintenance and failure to study who bears institutional costs and who can change the terms.
Existing accounts of path dependence, institutional lock-in, adaptive preference and non-domination help explain why established arrangements persist and why people come to accept them. Roberto Unger's False Necessity (1987) made the nearest argument: arrangements that look inevitable were made and can be remade, and institutions can be built to ease their own revision. The question here is related and distinct: how successful institutional investment affects other people's independent capacity to develop and control alternatives. I offer arguments and cases, not one mechanism established across every domain.
The research page has the field cases. Ethotechnics publishes proposed design standards and record formats.
Any proposed remedy also has to account for who will do its repair work, whose resources it uses, and how the people affected can challenge it. Necessary commitments can close some options while making others possible; their costs and authority need justification.
Whose ideals this puts under pressure
The argument presses on institutions and traditions that claim to defend freedom, independence, equality, merit, recognition or critical inquiry while leaving people dependent on powerful gatekeepers for the practical exercise of those ideals. Its force is the contradiction between what an institution promises and the conditions under which someone can actually use the promise.
- Liberal institutions. Equal rights can coexist with economic and professional dependencies that constrain their exercise. You Were Free to Choose.
- Market liberalism and meritocracy. Choice, opportunity and independence through competition depend on bargaining power.
- Universities and academia. Research and professional recognition depend on disciplinary approval and funding. Student Debt Does Part of the Deciding.
- Philanthropy and foundations. Funders influence which problems are investigated and which solutions receive support. The Boards That Decide What Counts as Possible.
- Progressive and reform organizations. Participation can expand without transferring authority to challenge institutional purposes. Consensus Without Conflict Is a Lie.
- Media and independent platforms. Dependence on owners, advertisers, subscribers or audiences shapes the boundaries of acceptable criticism. The News Is Paid to Keep You Anxious.
- Radical and counterinstitutional communities. Alternative communities can reproduce hierarchy and informal gatekeeping, and a coalition can be wrong about the people it speaks for. Everyone Institutions Wear Out Could Be One Coalition.
- Success and professional standing. Promotion and inclusion can raise the cost of criticizing the arrangement that produced them. You may never know how free you are until you say something that puts your belonging at risk. A Promotion Can Make Criticism More Expensive.
What this unsettles is not freedom, democracy, intellectual independence or equality as ideals. It unsettles the authority of an institution to declare those ideals fulfilled without examining whether the people subject to its power can independently contest that declaration. An institution cannot show it protects dissent by pointing to the dissent it permits.
Where to go next
Start here
The short guide to the framework: the core claim and the main patterns.
Method
The seven framework tests and the diagrams that show them.
Patterns
The recurring patterns, each with its questions, mechanisms, and reading order.
The archive
Essays tracing these patterns in healthcare, work, bureaucracy, platforms, and public administration, case by case.
Domain applications
The framework applied to healthcare, work, AI systems, education, disability and mental health, law, and politics.