Ministry of AI · Dispatch from 2047
Contribution Without Obligation: The Free-Rider Chapter
Editorial note. The Ministry of AI is a work of disciplined foresight: it describes the year 2047 in the present tense, and treats our own era as history. The institutions are imagined. The economics, the evidence and the historical parallels are real and sourced.
The hardest sentence I ever had to defend was not about money. It was four words a legislator said to me in a corridor in the seventh year, without hostility, almost apologetically: “But some will loaf.”
He was right. Some do. I have watched it in my own street, and I am not going to pretend otherwise, because the argument for the dividend does not need a fictional population of tireless volunteers to stand up. It needs an honest accounting of what unconditionality costs and what conditionality costs, and the willingness to publish both numbers next to each other.
The wound underneath the objection
Before I defend the design, I want to be fair to the objection, because it is not greed. It is reciprocity, which is one of the more decent things about people.
My mother spent nineteen years reading insurance claim files. She was good at it in a way that never appeared on any performance dashboard: she could tell, from the shape of a sentence, when a claimant was frightened rather than lying. When a system trained partly on her own decisions absorbed that judgment in the 2030s, she was fifty-six, and the thing that hurt her was not the loss of income. It was the loss of standing. She had been someone who carried a share. Now she was to be someone who received one.
The first dividend cheque did not fix that. What fixed it, slowly, was the Tuesday she started sitting with families arguing with the health service — unpaid, unrostered, nobody’s job. And here is the part that matters for this chapter: she wanted that recorded. Not rewarded. Recorded. When the Ministry consulted on the Contribution Record, the demand for it came overwhelmingly from people like her, who had lost a payslip and with it the only public proof that they were participating in the world.
Which is exactly why the Record is the most dangerous instrument the Ministry owns.
Why the record is never a condition
The moment you can measure contribution, someone will propose paying only the contributors. It sounds like justice. It is also, on the evidence, the fastest known method of removing money from the people who need it most.
The pre-Ministry safety net ran this experiment for decades, and the results are not ambiguous. When work requirements were applied to food assistance in the United States, exits from the programme rose by 23 percentage points — a 64 percent increase — among people already receiving it within eighteen months, and overall participation among adults subject to the requirement fell by 53 percent. Homeless adults were disproportionately screened out. And the employment effect, the entire justification for the policy, was not there: no measurable increase in work. Arkansas’s Medicaid work requirement produced the same shape of result, and the reviews of it years later confirmed that coverage fell without employment rising.
Read those numbers as an engineer rather than a moralist. A condition is a filter, and filters have a false-negative rate. This filter removed roughly half the eligible population to catch a group of non-workers who, when studied, turned out mostly to be working already, or sick, or unable to navigate a reporting portal. Even without conditions, take-up of benefits has never been complete; the economics of take-up has documented for decades that stigma, complexity and information costs keep eligible people out. Conditions add a second, larger leak to a bucket that already leaks.
So the Ministry adopted what is now called the read-only rule. The Contribution Record can be read by the person it describes. It cannot be queried by the Dividend Schedule, cannot be attached to an eligibility decision, cannot be produced in a hearing about a payment. This is architecture, not policy, because a record that could be used as a condition would eventually be used as one, in some fourth year, by some minister with a shortfall.
The obligation test
Every few years, someone proposes attaching a condition anyway. The Ministry does not refuse on principle. It applies the obligation test, four questions in fixed order:
- Does it raise employment in evidence, not in theory? Cite measured effects from an implemented condition, not a projection.
- What is its exclusion cost line? The estimated number of eligible people the condition removes, published in the same document as the expected saving.
- Who is screened out first? Conditions are not neutral filters. They sort by paperwork capacity, which sorts by illness, housing status, language and age.
- Does the same result exist without a condition? If voluntary provision achieves it, the condition is buying compliance, not outcomes.
In forty years, two proposals have passed the first question. None has passed all four. The test is boring on purpose. Boring is what stops a moral intuition from becoming an eligibility system.
| Conditional assistance (pre-2030s) | The dividend with a read-only record | |
|---|---|---|
| Basis of the claim | Need, proven repeatedly | Inheritance, held permanently |
| What the state verifies | The person’s compliance | The machine’s yield |
| Failure mode | Eligible people excluded — participation fell 53% under work requirements | Some recipients work less — about 1–2 hours per week |
| Who is hurt by the failure | The sickest and least-housed | The public purse, measurably |
| Administrative cost | Caseworkers, audits, appeals, portals | A payment file |
| Effect on standing | Applicant | Owner |
The right-hand column is not a utopia. It is a different, smaller, better-documented set of costs.
Conceding the free-rider case properly
Here is the number I owe the legislator in the corridor.
When guaranteed income was tested at scale in the 2020s, recipients did reduce their work. The largest randomised study found a decline of roughly one to two hours per week, a fall of about four percentage points in labour-market participation, and around $1,800 less in annual individual earnings. The largest single increase in time use was leisure. Not caregiving, not study — leisure. Advocates at the time preferred not to lead with that finding. The Ministry leads with it, in the annual report, on the same page as the yield figures.
Because the honest version of this argument is not nobody loafs. It is: loafing is a priced, published, tolerable line item, and enforcement is a more expensive line item that hits the wrong people. A payment that removes a hundred thousand sick applicants to deter ten thousand idlers has not defended reciprocity. It has bought a feeling with other people’s rent.
There is also a claim underneath the pricing. Reciprocity assumes an unpaid debt. But the intelligence generating the yield is a compression of the human record — the books, the forums, the repair manuals, the publicly funded science, the publicly educated engineers. The dividend is the return flowing back to the authors of that inheritance. My mother is not receiving a favour that obliges her to prove herself. She is receiving a distribution on something she and everyone before her helped write. Capacity is not contribution, and work that must be done to survive is not chosen work.
What it looks like when it works
My middle son is fifteen and spends three afternoons a week rebuilding a boat with two retired men who used to fit windows. Nobody is paid. Nobody is inspected. The Record notes it, because he asked it to, in the same way people once kept a diary — proof to himself that the hours went somewhere.
If the obligation test had failed, and the Record were an eligibility instrument, that boat would still be built. But the two retired men would be filing hours, my son would be a dependant with a compliance obligation, and someone in an office would hold the power to decide that boats do not count. The difference between those two worlds is not the boat. It is who is permitted to define what a life is for.
Where the design fails
Three admissions.
The first is the one above: measured work reduction, real and paid for, and the largest chunk of the released time is leisure rather than care or study. If you believe leisure is worthless, this design is not for you. I think a society that has automated the smart work and cannot tolerate rest has not understood what it built.
The second is that voluntary recording is self-selecting. The Contribution Record over-represents the confident. The people whose contribution is least visible — the ones caring at three in the morning — are still the least likely to log it, which means the Record understates precisely the group it was built to honour.
The third is political. Unconditionality is defensible on evidence and vulnerable in a bad year, because some will loaf is four words and the rebuttal is a table. Every generation has to relearn that the cheap sentence is not the true one. We have kept the promise for forty years. I do not assume the forty-first is free.
FAQ
What does contribution without obligation mean? The Ministry records what people choose to do with their time and never uses that record to decide whether they are paid. Contribution is honoured; it is not a toll gate.
Doesn’t an unconditional dividend create free riders? Some. The measured effect is a reduction of roughly one to two hours of work per week and about $1,800 in annual earnings among recipients in the largest randomised study of guaranteed income. That is the cost, published, not denied.
Why not attach a work requirement? Because implemented work requirements cut programme participation by more than half and raised exits among existing participants by 23 percentage points without raising employment. They screened out the homeless first. They did not create workers.
What is the read-only rule? The Contribution Record is legible to the person it describes and to no one with power over the payment. It cannot be queried by the Dividend Schedule or used in an eligibility decision.
What is the exclusion cost line? The published estimate of how many eligible people a proposed condition would remove, printed beside the saving it promises. Most conditions look worse once both numbers appear on the same page.
Is contribution still valued in 2047? More than it was when it was compulsory. Removing the requirement did not end participation; it ended the pretence that participation performed under threat of hunger was ever freely given.