The answer already exists. It is filed under someone else’s problem.

Why this exists

People are good at analogies. They are bad at finding them.

In 1980, psychologists gave students a hard problem: destroy a tumour with radiation without burning the healthy tissue around it. About one in ten solved it.

A second group read a story first. A general attacks a fortress by splitting his army into small groups that converge from different roads. Same problem, different nouns. Solving rates roughly tripled.

A third group was told to use the story. Three quarters solved it.

The gap between the second and third groups is the whole finding. People who had just read the answer mostly did not think to apply it. What failed was not reasoning. It was noticing that a relevant case existed.

Gick & Holyoak, Cognitive Psychology, 1980 and 1983.

What we found looking at real cases

Transfers happen when one person happens to know two fields.

We took five documented cases where a solution moved from one domain to another and looked at how each one actually happened.

Cure Violence. Gary Slutkin spent a decade running epidemic response for the WHO, came back to Chicago, and was asked about street violence. He looked at the maps and saw an epidemic curve.

Camden hotspotting. Jeffrey Brenner was a Camden physician who sat on a police reform committee, saw CompStat crime maps, and asked whether health data would look the same. It did.

CitiStat. Martin O’Malley saw CompStat in New York and brought it to Baltimore city agencies. He did not just copy the idea. He hired Jack Maple, one of its creators, as a consultant.

Forensic ACT. A psychiatrist treating patients who cycled through jail knew both the clinical model and the courts.

Four of five needed one person carrying two fields in their own head. The fifth is the one that failed: Mexico’s conditional cash transfer programme was studied carefully and brought to New York in 2007 by a policy team scanning international evidence. No carrier. It underperformed.

So the current mechanism for connecting fields is biographical accident. A transfer waits for someone to happen to have had the right career. That is what a database can replace. Not insight. Coverage.

The verified case

Two records the system had never been told were related.

Each record in the corpus carries two structural tags: the mechanism it uses, and the institutional failure it corrects. Those tags are assigned from the source paper, one record at a time, with no knowledge of any other record. The match below was computed from the tags alone. It ranked second out of 1,948 candidate pairs.

Where this stands

Four confirmed transfers, and one the field reached first.

Twelve problem-to-solution candidates the system proposed were checked against published literature. Four are documented transfers with the lineage stated in the sources. Three of those four also address the problem they were matched to.

Baltimore’s CitiStat took CompStat from the New York police department into municipal services in 1999. The mayor hired the man who built CompStat as a consultant. Performance management moved from public administration into schools, and motivational interviewing moved from addiction counselling into managing high blood pressure. In each case the source literature names where the method came from.

Two of those three are among the best-documented cases in their fields. That matters: it shows the system can surface a real transfer without being told about it, and it does not yet show the system can surface one nobody has written up.

The fourth confirmed transfer is participatory budgeting, which moved from Brazilian city government into schools and is documented working. It does not clearly answer the problem it was matched to, so it counts as a real transfer and not as a real match. Of the rest, four were not transfers at all, and three were correct matches inside a single field. We keep all of them, because a pattern in what gets rejected is how the scoring gets corrected.

The candidate we would point at is not on that list. Federal homelessness funding requires agencies in a region to coordinate as a condition of the money. The system matched that against fragmented care for sickle cell disease, where specialists work in isolation. Then we checked, and Congress had funded that same structure in 2004. Nobody told the system that. Arriving independently at something the field already reached is a stronger result than an untried idea would have been, and it is one candidate in twelve.

The open proposal

What would have to be true.

Nobody proposed this pairing. The system matched a homelessness funding rule against fragmented sickle cell care because both describe the same structure: providers who each hold part of a person’s situation and no body that holds the whole of it. Then we checked, and found the pairing was not new. Congress created a Sickle Cell Disease Treatment Demonstration Program in 2004, and since 2006 it has funded regional collaborative networks on exactly this logic. Nobody told the system that. It arrived at a structure the field had already reached.

The mechanism has two halves. Federal homelessness funding requires every agency in a region to feed one shared registry, and it requires them to join one coordinating body. The registry is the cheaper half.

That is the gap the sickle cell literature names. Adults leaving paediatric care face a sevenfold rise in mortality, largely because there are almost no adult providers who specialise in the disease, so patients scatter across primary care and emergency departments. New York City’s health department named the absence of a comprehensive patient registry as a core failure last year. The information exists. It sits in pieces nobody can assemble.

Something has already been tried here. Between 2018 and 2019, North Carolina attempted to push evidence-based sickle cell guidance through the state’s Medicaid managed care network, a network covering roughly ninety percent of the state’s primary care providers, and directly to emergency department staff at a smaller scale. Utilisation did not move: no significant change in specialty visits, emergency visits, or hospitalisations.

The authors say something more useful than the null result. They could not confirm the guidance reached the providers at all. Delivery ran through quality-improvement specialists and the study reports being unable to track whether any given practice received it. So the finding is not that knowledge did not help. It is that nobody could establish the knowledge arrived, which is the same shape as the problem: information exists, and no body holds it or moves it.

That does not test the funding condition, which is a different mechanism from guidance. It sets a bar: whatever is tried next has to be something whose arrival can be verified.

The two approaches came from different places, and that is the part worth noticing. North Carolina’s came from inside the field. When someone who knows the disease sees that providers lack knowledge, distributing knowledge is the obvious move. Ours came from a housing funding rule, which nobody inside haematology would have thought to look at, because there is no reason to.

That discovery turns the question around. It is no longer whether regional coordination would help. It is why the coordination that exists has not moved the outcome, and the answer names what the transfer would actually need.

Scale is the first part. The federal sickle cell treatment programme runs on about $8.2 million a year against $4.04 billion for the homelessness competition. That is roughly eighty-two dollars per patient per year, which cannot compel anything.

But the deeper condition is that coordination needs somewhere to route people. Coordinated entry works in homelessness because permanent supportive housing units exist to assign. Adult sickle cell care has no equivalent, and the reimbursement structure actively prevents one: a dedicated sickle cell unit cut admissions by 43% and likely lost its institution money under how inpatient care is paid for, while oncology procedures pay better than non-malignant blood disorders, so the specialists migrate away. Coordination cannot allocate capacity that does not exist.

What would change that reading is a region where adult specialty capacity demonstrably exists and is underused while outcomes stay poor. That would make routing the binding constraint rather than supply. We have not found one.

What it would mean

The same barrier, in four buildings on the same street.

A housing office turns people away because the paperwork defeats them. A clinic loses patients between referral and first appointment. An election office watches eligible voters fail to register. A retirement plan sits unused because enrolling takes an afternoon.

Four institutions, four literatures, one structure. Each has partial answers the others have never read.

That is not a knowledge problem. Every one of those fields has done the research. It is a retrieval problem, and retrieval is what a corpus is for.

The working version of this is unremarkable to describe. Someone facing a problem writes it in their own words. They see which other fields have the same structure, what each of them tried, what it required, and where it broke. Then they decide, because deciding was never the part that was missing.

This is a draft of the argument, not a description of a finished product. Four transfers have been verified against literature, out of twelve candidates checked. The rest is a method that produces candidates and a corpus too young to have met practice.