You Pay to Be Heard, and the Worst Off Pay Most

An institution that listens through surveys, appeals, and phone lines makes the people it serves pay to be heard.

Ask a hospital how its patients are coping and the answer is a survey, which takes forty minutes from someone who is already ill. An insurer asked whether claims are going well offers an appeals process, which costs an afternoon for someone who has already lost the month. A bureaucracy asked whether it’s accessible offers a phone line, staffed during the hours the person calling is at work. None of these are free. Each is a toll, paid in time and energy at the moment the person has the least of both.

Under all three sits the same simple mechanism, and it’s almost never said out loud. Measuring human burden through voluntary input asks the person carrying the burden to do one more unit of unpaid labor for the organization responsible for it.

A shorter form leaves the mechanism untouched: the people who reply are still the people who had something left to spend.

Respondents divide along that line. People with a thin surplus of time, literacy, and composure file the complaint or work the second-tier appeal. People who are more thoroughly depleted often stop. The claim goes unfiled, the grievance gets dropped, the referral is never made, and some leave the system entirely rather than keep spending. The institution reviews a sparse log, sees few complaints and mostly resolved tickets, and concludes that things are working.1

A sensor that depends on the victim’s remaining energy is broken rather than merely bad. Serious harm makes people louder, but having less to spend makes them quieter, so the reading can’t tell you how bad things are.2 So a quiet feedback log is consistent with a failing system and can easily be filed as evidence of a working one. Silence alone proves neither: it has to be read against the trail of abandonment, unresolved need, sampled experience, and who can reach the channel at all. Assume the People in Charge Already Know shows what happens to an institution whose signals are rich and whose repair is unmotivated: the telemetry arrives and feeds tolerance bands, pricing adjustments, and symbolic response loops instead of anything that changes a workflow. The inverse case is worse: the signals are thin because much of the population with the least to give has already left the channel.

An institution should read how much friction the system generates, which is already on the record and doesn’t require anyone’s consent to compile. Whether people report distress is the wrong reading, because once reporting is itself expensive they won’t, reliably or honestly.

The trail left behind

An institution doesn’t need to ask whether people are being ground down, because the interaction itself already left a trail. Someone who gives up on a form leaves abandonment at a specific step, a repeat of a login, a re-upload of a document they already sent, a session that ends without submission. The only decision left is whether to read it as noise or as a fault report.

What gets measured is a delta between the path the process advertises and the path people actually walk. In software these get treated as funnel drop-offs. That’s why software teams catch them, and why the same institutions that write the software treat their own forms as user attrition.

That trail is frictional exhaustion, and it reads straight off the form. Off-hours spillover between 8 PM and 6 AM and compensatory workarounds are ambient readings too, and they already have instruments in The Unpaid Fixes Are Already in Your Logs. Measure the Machine, Not the Worker sorts out which of these readings stay a measure of the system once a manager can rank people by them.

Stop before the wire burns

Site reliability engineering has run outage budgets and circuit breakers for two decades. The Front Line Looks Like a Cost Until It’s Gone borrows the idea to rotate people out of boundary roles on a schedule, before they ask, the way a service sheds load so no single node takes the whole surge. The inverse rule turns the analogy into an instrument.

Every component absorbs a surge right up until it stops being able to, and a fuse has no opinion about the wire’s feelings. The wire opens when the load passes a threshold somebody wrote down in advance. The institutional equivalent is a hard cutoff on human load, applied automatically and without consultation:

• When the patient-to-nurse ratio crosses its agreed ceiling, intake stops until it recovers.
• When open cases per social worker pass the limit, new intake diverts elsewhere rather than waits.
• When a queue exceeds its service window, work is shed on a published priority order rather than soaked up in silence until someone breaks.

What a hospital or agency calls resilience is usually staff carrying load past the point where intake should have stopped. A stoppage is a different event, and it’s worth what it costs: visible, expensive, and attributable to whoever owns the process. Endurance is cheap to everyone except the person supplying it, which is why nobody is invoiced for it.

Make friction cost the institution

Institutions externalize friction because delay, opacity, and complexity are free to the institution and expensive to the person. Reversing that is the cheapest available lever, because it builds the sensor out of existing accounting rather than out of goodwill.

• Administrative tolling means a deadline that resolves in the claimant’s favor. If an appeal, prior authorization, or benefit application isn’t adjudicated inside a human-scaled window, the decision defaults to approved. The decision guarantee in “We Believe You” Is Not a Decision forces a ruling inside the window; tolling also decides which way a missed window falls.
• Liability on the clock puts a price on the wait. When institutional delay carries a direct financial or regulatory consequence, latency stops being an unrecorded externality and becomes a line item with an owner. Institutions tend to build their own instrumentation once delay threatens their own balance sheet, because at that point the signal is already priced.

Reliability engineering already knows how to cap a scarce resource and then watch the cap. What it has never had to do is hold a human being to a latency budget, because humans don’t return an error code.

• Write the budget before shipping the mandate. If a workflow, form, or compliance rule is going to exist, the sponsoring body has to state in advance a ceiling on the labor it will cost. That could be a cap on total cognitive minutes for 95% of eligible applicants. A process whose labor cost was never written down has no argument to lose. The ceiling line of the remainder statement in The Rule Works. She Doesn’t Fit. does the same work for dropped cases.
• Treat the breach as an outage rather than a variance. When the observed median comes in at three times the budget, the process is degraded back to a simpler baseline while someone fixes it. Processes don’t get to be merely inconvenient on a schedule, because the median person has no schedule for recovering.

Separate stewardship from those who carry the load

Auditing burden can’t belong to the same people the burden is landing on, and it cannot be a complaint queue that waits to be filled.

• Dedicated friction auditors go through the system as an ordinary user, read the telemetry, and carry binding authority to deprecate the workflows they find broken.
• Collective rather than individual redress treats a defect as a structural fact. If a notice or a decision is found to be unlawful or unreadable, the remedy is structural and retroactive: recall, correct, and re-adjudicate every downstream case automatically. An institution that repairs a defect only through individual appeals has turned its own error into more of the labor that caused the harm. One Claimant Won While 70,000 Others Stayed Underpaid argues for answering the pattern at the size of the rule.

What the sensors will be routed around

Every sensor here can be routed around, and the work will get moved to wherever the sensor isn’t looking: onto contractors, or onto the patient as a second phone call. A Great Metric Can Mean Someone Is Quietly Fixing It treats that as the normal outcome rather than a loophole, so the design has to start from it.

Off-hours telemetry is the sharpest double edge. It’s the cleanest available signal of uncompensated labor and the easiest to defeat without showing. The workaround is a schedule: batch the charting at lunch, close the record after hours, finish the queue from home. The signal holds up only as far as staff trust the institution with it. That dependence means the signal measures institutional trustworthiness while being reported as a measure of burden.

And capacity throttles can be defeated without anyone noticing, because a queue that stops taking new work looks identical to a queue that has stopped existing. Automatic intake diversion is only honest if the diverted demand stays visible in the institution’s own reporting. That means the circuit breaker needs to be metered on the far side.

So the architecture holds only with a standing expectation that absorption will be rerouted rather than eliminated, and that the rerouting is the finding. That reverses the burden of proof in a way most institutional metrics never do: the question moves from whether people are coping to where the work went. The sensors are how you get it to show up on the record. None of them makes it stop being someone’s job to find out.

Whoever sets the ceiling has already decided what counts as too much. That’s why Measure the Machine, Not the Worker argues that the people being scored must be able to contest the categories, the weights and the ceiling.

Where the loop closes

An extractive institution and a corrigible one send the same failure to different places:

An institution that treats human adaptability as free, unlimited, and off the books has written no limit into its operations, so nothing in its records shows where its loop closes. Build the limits in (the cap, the default, the deadline, the stop), and the question changes. It stops being how to make people endure this and becomes what has to be rebuilt so the whole arrangement fits inside the boundaries of one finite life.