Belgium works well as a set of institutions. It is thinning in its culture and shared meaning.
The base of life is well cared for. Safety, welfare, and fair procedure all rest on an old machine that still runs, and this paper says so plainly. Above that base the picture changes. The middle is hollowing out: the forms are still there, but less and less passes through them. The country is very good at noticing what goes wrong and very poor at fixing it. The higher things people want — a sense of aliveness, of things fitting together, of shared direction — have mostly become things you buy, or achieve alone. The one place where an ordinary Belgian can still rise by giving something, rather than by consuming, is the club and the association. The highest place of all — a shared sense of what the whole country is for — looks empty at first. It is not. It is held quietly by the rules-and-rights order itself, and the cost of holding it that way is passed to the future. Where the state and the old institutions leave a need unmet, others step in. On present trends, the ones that win are the easiest to reach, not the ones that last.
This reading does not come from one analyst's opinion. It comes from a structural model of how societies hold together, described in section 2. To check that the reading can be repeated, two independent readers applied the model to Belgium on their own — with no contact, and without seeing each other's work. Where they reached the same result, we report it as a finding. Where they differed, we say so; the differences taught us as much as the agreements. The readers are graders. The model is the source. What the readers are, and are not, is set out in section 2.
A one-page Dutch and French summary is at the end of the paper.
Every measurement needs a standard. Most measures of social health borrow one without saying so — from economics (how many people take part), from psychology (how satisfied people say they are), from politics (how much they trust institutions). This paper states its standard openly. It is one that a reader of any belief can accept before seeing what follows.
A society exists to do two things.
First, it feeds the floors. People need real access to each level of what human beings seek, in rising order: safety of the body; real belonging between people; fairness and clarity; vitality — being alive, growing, intelligence in action; integration — things working together as a whole, deep peace; and direction — the whole pointing somewhere, together. These are not luxuries stacked on top of needs. Each level is a real need. And a need that goes unmet does not sit still. It goes looking for something to hold on to.
Second, it opens the ladder. Here the standard makes its one demanding claim: the higher levels cannot be handed to a person the way safety can. They can only be reached by taking part in the whole — by giving to it, not by consuming from it. A society that offers its members only the role of consumer at the top — belonging as a spectator, well-being as a purchase, integration as a product — is failing this test even when its satisfaction surveys look fine. And the whole includes the natural world: care for creation is part of the standard, not an extra.
One objection deserves an answer before any finding, because it is the strongest one. A liberal reader will say that having no shared final purpose is not a fault but an achievement — that a diverse society is held together exactly by procedures, rights and mutual restraint, and that what this paper calls an empty place is the point.
The answer here is structural, not moral. Anything that stays one system, rather than falling into many, has to be held together by something. And what holds it has to be strong enough to carry the load that its own size and complexity create. This is not a claim about religion. It is close to a rule from cybernetics — a system that controls something needs at least as much variety as the thing it controls (Ashby's law) — and to Durkheim's observation that as a society's division of labour deepens, the form of what binds it must change, or the society drifts into what he called anomie, a loss of shared norms. Different words, one point: the power to bind has to grow as complexity grows. A society this varied, this mobile, this open to strangers and to speed carries far more load than the arrangements it inherited were built to hold. So the things that bind — traditions, ideologies, procedures, institutions — are not all the same. Some can carry more load than others. And that difference can, in principle, be measured.
This is why proceduralism is not dismissed in what follows. It is treated as what it is: a real holder of the top place, with real achievements, measured on the same terms as every other candidate. The question is not whether it is right. The question is how much it can carry, who keeps it alive, and where its costs land.
What the model looks at. The model maps a society as a set of parts and the bonds between them. The parts are the things a society needs someone to do, the players who do them, and the levels of need people are trying to meet. The bonds are the relationships that carry a society: family to family, worker to employer, citizen to state, member to club, and so on. It then reads each part and each bond. For each part: is it working, hollow, or turned against the whole? For each bond: does it carry meaning, or exchange, or shared structure — one of these, or all three?
A wide reading, not a short checklist. The Belgium reading holds, in one connected picture: fourteen things a society must be able to do; six levels of need; around twenty load-bearing bonds; and a dozen or more active players — each scored for Flanders and Wallonia separately, with Brussels noted where it differs. That is well over two hundred separate judgements, held together in a single view. A figure like “the two readers agreed on 18 of 28” (below) refers to one sheet inside this picture. It is not the whole of what the model measures.
What is different about it. Most measures score how well a society performs on a list of outcomes. This model measures something else: how the society holds together — whether the bonds that carry it are whole or hollow, and how the parts relate to one another. That is the thing that gives way first under stress, and it is exactly what an outcome measure cannot see. The counts of members, of trust, of volunteering can all hold steady while every bond beneath them drops from carrying three things to carrying one.
It can also be run forward. Because the model represents a society as parts and bonds, the same structure can, in principle, be run as a simulation — not to predict events, but to show which arrangements can hold and which cannot. This paper does not run that simulation. It reports a single reading, at one point in time. But the reading is built in a form a simulation could later use.
What the readers are. This is a modelling exercise. The model is a way of describing structure. The two-reader run checks one thing only: that a person other than the model's author can apply it and reach the same description. It is not a test of a hypothesis. Two readers — AI systems from two different companies — each received only the model, and worked blind: no access to our earlier work, to each other, or to any expected answer. Their sheets were sealed before we compared them. They are not evidence about Belgium. They are careful, consistent graders, fast and cheap enough to run again and again. What their agreement measures is whether the model is usable — not whether its output is true.
Both readers, on their own, scored the bottom of the ladder as the country's real strength. This deserves saying without irony, because everything else in this paper stands on it.
Safety is real. Order is kept, violence is low by any world standard, and the police carry broad trust. Material welfare is real. The social-security system links almost every resident through insurance, pensions and care, and it held through a pandemic, floods and an energy shock. Fair procedure is real. Rights are predictable, courts are independent, and the constitutional order is respected across political camps. And detection is real. Belgium reliably notices what goes wrong. Its press, its parliamentary inquiries and its audit bodies bring failures to light with a consistency many countries would envy.
This is what a century and a half of institution-building bought — much of it built by the very pillars whose loss this paper will soon describe. The machine works. That is exactly why the quieter losses go unseen. A country whose floors are this well fed does not feel itself thinning higher up, because the thinning happens above the level the usual measures watch.
The middle of the ladder is where both readers began to write the word hollow — the form is present, the function leaks away.
Care still runs, but more and more through paid services, stretched over the gaps where thick family and neighbourhood webs used to be. Childcare and elderly care are visibly strained, and the volunteers who carry local life are ageing faster than they are replaced. Passing things on — knowledge, skill, shared meaning — to the next generation was scored hollow in both school systems: the machinery still teaches, but what it once carried through the teaching has faded. Justice still decides, but so slowly that the sense of fairness leaks away in the waiting. Conflict is settled — in law, in money — rather than repaired. There is no longer a widely available way to actually close a wound between people or groups.
And then there is the one cell where the two readers, working blind, wrote almost the same sentence. Belgium detects its failures superbly — and does not correct them. Inquiry reports are shelved. Reforms stall for decades. The same problems are diagnosed again by each new government. One reader called it “the system's known broken link.” The other wrote “recognition often exceeds implementation.” In the model's terms: the power to detect is healthy, the power to correct is hollow, and the power to keep the fixing skill alive across generations — the supply of teachers, judges and civil servants who carry it — is thinning in both regions.
The reason was described by Belgium's own sociologists decades ago. The pillars — Catholic, socialist, liberal — were not only political blocs. Each was a braid. A single membership carried meaning, welfare, schooling, mutual aid, politics and friendship at once. One relationship, many strands. When the pillars fell, the strands did not vanish — they came apart. Exchange went to the market. Structure went to the state. Meaning went private. Today's Belgium keeps most of the single strands and has lost the braid. Both readers, scoring the country's main bonds, found the same pattern: the thick bonds (citizen–state, worker–employer) carry money and law but no longer meaning; the meaning-bonds (family, club, identity) carry no structure. The result is a society that is well run and culturally coming apart at the same time — which is just how Belgium tends to describe itself, usually without being able to say why.
First, a word this section needs. A society needs certain jobs held by someone — the job of keeping people safe, the job of settling disputes, and so on. This paper calls each such job a “place” (or slot). The top place is the highest one: the job of holding, and keeping alive, a shared sense of what the whole society is for — its direction. In a religious age the churches held it. The question now is who holds it, if anyone.
Above the middle, the reading finds three things that are not just thinning but missing, or wrongly held.
Every society once had people whose role was not advice but integration — helping a person or a community fit the pieces of a life into a whole. Both readers scored this at or near absent. What fills the space is a paid market of therapy, coaching and wellness: real relief, one person at a time, but advice without direction — integration sold by the unit. It is the difference between a guide and a pharmacy.
The model asks, at every level, whether an ordinary member can rise by giving through what is on offer, or can only receive from it. The answer, agreed by both readers for both regions, is stark. Belonging — the club, the association, the volunteer brigade — is the only level where giving is still the normal path. Safety is delivered to you. Fairness is administered for you. Vitality is bought or performed. Integration is purchased. Belgium still has, in its association life, one of Europe's best machines for letting people contribute — and it has quietly shut that machine into a single floor of the building.
Here the two readers seemed to disagree, and the disagreement turned out to be the most useful result of the whole exercise. One reader scored the top place as empty for most people: the church's forms survive for a shrinking minority and at funerals, and nothing has replaced them. But the model contains a rule: a reader who finds a place empty must ask what is invisibly carrying the weight, because a need at a load-bearing level does not sit still. Following that rule, the first reader pointed at the welfare state itself acting as an unspoken final purpose. The second reader, working blind, simply named the holder directly: human rights, liberal democracy and pluralism, held as the highest public ideal. Two routes, one finding: Belgium's top place is not empty. It is held, without anyone saying so, by the rules-and-rights order — peace, rights and coexistence as the final horizon. This is a real holder, with real achievements, and it is scored here on exactly the same terms as every other holder in the study, not a stricter one. But it is a holder of a particular kind. It organises coexistence without pointing anywhere. It is inherited, not actively renewed. And — as the fate of every earlier unspoken holder suggests — a place held by a form that no longer says why is a place that stronger-flavoured bidders will, in time, contest.
The standard includes the natural world, and the reading scored Belgium's carrying of it hollow in both regions: there is active regulation, and there are real movements at the edges, but no living, widely-shared carrier of balance with nature. The nitrogen crisis in Flanders and the slow reckoning after the Wallonia floods each showed the same thing in their own way — rules colliding with life, with no shared practice to make the balance mean something rather than merely be enforced.
The differences between the regions matter. Flanders holds the country's densest machine for contributing (its association world) and its sharpest identity politics. Wallonia holds warmer local life, but both readers scored its material floor as not merely dependent on transfers but borrowed — it would sag if the transfers stopped — and its vitality thinner still, with a jobless share for whom the ladder offers no way to contribute at all; its fairness place is contested between an old machine and a live challenger. So the gap between the two regions is structural, not just warmth against wealth: Flanders thins from a higher floor than the one Wallonia stands on. Brussels differs from both: its safety floor is under visible strain, its coordination has been left unmanned for long stretches, and it is the one place where the top place has living mass carriers — for a large minority — while the wider population runs on parallel ladders that barely share a bond. Brussels is not an exception to this reading. It is its preview.
Needs the official institutions leave unmet do not stay unmet. The last part of the model asks who is actually bidding for the open places, and what each bid costs.
The largest bidder, by reach, is the market — not as villain but as mechanism. It bids at almost every level: safety as insurance and assets, belonging as consumer tribes and festivals, vitality as career and lifestyle, integration as the wellness industry. One reader, knowing nothing of this project's theory, explained its success in a line we could not improve: it wins “because it provides immediately accessible substitutes.” That is the whole cost-logic of the contest. The easiest substitute wins the empty place, not the most lasting one — and it pays for that ease later, in the exhaustion of the person holding it. The burnout and long-term-sickness figures are, in this reading, the receipt.
The fastest-growing bidders are the identity movements, in both regions: belonging, a sense of grievance-as-fairness, and a kind of unity, offered in one package. The model prices the package: real warmth inside, costs sent to named outsiders, and a direction that stops at the group. They are winning where they win because they are the only mass bidders speaking to belonging and fairness in the language of meaning since the pillars fell silent. The need they answer is real, even where the answer is not.
The ledger — asked of Flemish, Walloon and federal government separately, and of the EU where it is the real actor — came back much the same from both readers.
Government contributes the floors and the coordination — genuinely, ably, at scale. Nothing in this paper suggests otherwise.
Government substitutes where it should enable. The old civil society was not replaced by the state; it was bought by it — professionalised, subsidised, put on contract — until, in one reader's words, the state “bought the native carriers and now owns their payroll.” A simple test makes it concrete: if the state stepped out of a place, is there anything underneath? In Flanders, sometimes — the youth movements, the volunteer world. In Wallonia the same test, both readers agreed, fails almost everywhere above the family.
And here the European record deserves one paragraph, named plainly. East Germany — the communist German state that existed from 1949 to 1990 — provided for its people, but did not let them build and run their own institutions. When it fell, the society left behind had been cared for but not formed from within. The former East is today the part of Germany where support for the far right is strongest. Providing everything while forming nothing keeps the present stable by using up the future's ability to stand on its own.
A clarification, because this is the point most easily misread. Nothing above is an argument for a smaller state. The state is essential as the scaffold of the upper levels: it guarantees the floors, enforces fair treatment, funds what communities cannot fund, and protects the people whom communities shut out. But a scaffold is built to hold a structure up, not to be the structure. What the ledger flags is the point where public power stops making sure carriers exist and becomes the only carrier — treating care, teaching and repair as services to be delivered — which turns members into service-users and wears out the staff who deliver it. The opposite mistake is just as real, and the model scores it too: carriers left entirely to themselves can be uneven, closed, and hard on people they do not count as their own. That failure is not outside the model — a bond that stops at its own group is exactly a bond that does not open outward. The path that survives both mistakes is shared production: public guarantee, community carriage.
Government caps the upper levels — mostly without meaning to. Neutrality rules keep meaning-carriers out of shared public space; funding rules admit civil society only as a service-deliverer. The effect, whatever the intent, is that the levels above fairness are officially nobody's business. This raises a question, put here honestly as a question: can a public order stay neutral about the top half of the ladder without thereby deciding it — because the reading suggests the place does not stay empty; it gets filled by whatever the rule does not see.
These are directions, not prescriptions — and none of them requires accepting this paper's theory. Each stands on the findings alone.
Short version — the full protocol, the two sealed reader sheets, the model text and the agreement tables are provided as a separate technical and replication appendix.
This reading was produced by a structural model with five parts (capability coverage; a census of what people hold on to; a reading of how bonds are braided; a government ledger; a map of who fills the gaps), applied by two blind, independent, sealed readers from different AI families, with agreement measured cell by cell. It follows an identical study of Germany, which passed the agreement levels set in advance; on this study one coverage level was missed, and the numbers are given plainly in section 2 and in the appendix. Every claim in this paper traces to a cell both readers scored, a disagreement analysed in the open, or a check against public trend data.
Two AI systems may agree because they were trained on overlapping text, not because they read the same country. Drawing them from different companies reduces that risk; it does not remove it, and we do not claim it does. Three things in the results argue against simple echo, without settling the matter. The disagreements were systematic, not random — the same five capabilities split the same way in both regions, which is the mark of a loosely-worded definition, not a shared prejudice. The study's most important finding (section 5) was produced by one reader following a written rule to a conclusion the other reached by a completely different route — which is not how a shared prejudice behaves. And the readings were checked afterwards against trend data neither reader was given. One limit belongs in the same place: for one of the two systems we could not confirm that no memory of earlier conversations was in play, so its independence rests on the absence of any sign of leakage in its sheet rather than on a guarantee. The decisive test is not more machines. It is human expert readers applying the same sheets — phase 2 below.
This is phase 1: checking that the model can be applied — that a reader who has never seen it can use it and reach the same cells as another such reader. Phase 2 is planned and not yet run: the same sheets, sealed and blind, read by a panel that mixes human experts with machine readers, with human-to-human and human-to-machine agreement reported separately. For Belgium the intended human panel deliberately pairs readers of opposite conviction — a secular counsellor and a religious counsellor among them — because the findings worth keeping are the ones that survive that pairing. Phase 3 is a check against existing data series (the European Values Study, the national statistics office, the regional well-being indicators): not to prove the reading right, but to find where structural scores and survey series move together, and where they part, and why.
The model itself is published in the technical appendix, so that any reader — a person or a machine — can repeat the Belgium reading and compare sheets. Ways this reading could be shown wrong, set out in advance, include: a rerun with expert human readers failing to reproduce the main findings; the empty places the closure rule flags failing to match observable stress, or failing to draw new fillers; and trend data moving against the readings over the coming years. This is a first reading, not a verdict. We publish it to be corrected.
België is institutioneel sterk en cultureel aan het verdunnen. De fundamenten — veiligheid, welvaart, procedurele rechtvaardigheid — worden gedragen door een geërfde machine die werkt. Daarboven verdunt het weefsel: zorg en overdracht hollen uit, het land detecteert zijn problemen uitstekend maar corrigeert ze niet, en de hogere niveaus van wat mensen zoeken zijn geprivatiseerd tot iets wat je koopt of alleen bereikt. De enige plek waar een gewone Belg nog kan stijgen door bij te dragen is de vereniging. De hoogste plek — de gedeelde richting — is niet leeg: ze wordt impliciet bezet door de procedurele orde zelf, met de rekening naar de toekomst. Waar noden onvervuld blijven, winnen de meest toegankelijke substituten — niet de meest duurzame. Dit is geen pleidooi tegen de neutraliteit van de staat, noch voor een kleinere staat: de staat blijft onmisbaar als draagstelling, maar een draagstelling is gebouwd om een structuur te dragen, niet om er zelf een te zijn. Vijf richtingen volgen uit de meting: financier eigen draagkracht in plaats van substitutie; open bijdrage-kanalen op elk niveau; investeer in vervlechting in plaats van losse strengen; geef rentmeesterschap een levende drager; en behandel de bovenplek eerlijk.
La Belgique est institutionnellement solide et culturellement en voie d'amincissement. Les fondations — sécurité, prospérité, équité procédurale — sont portées par une machine héritée qui fonctionne. Au-dessus, le tissu s'amincit : le soin et la transmission se creusent, le pays détecte remarquablement ses problèmes mais ne les corrige pas, et les niveaux supérieurs de ce que cherchent les gens ont été privatisés. Le seul endroit où un Belge ordinaire peut encore s'élever en contribuant est l'association. La place la plus haute — l'orientation partagée — n'est pas vide : elle est occupée implicitement par l'ordre procédural lui-même, la facture étant reportée sur l'avenir. Là où des besoins restent sans réponse, ce sont les substituts les plus accessibles qui gagnent — non les plus durables. Ce n'est un plaidoyer ni contre la neutralité de l'État ni pour un État plus petit : l'État reste indispensable comme échafaudage, mais un échafaudage est fait pour soutenir une structure, non pour en tenir lieu. Cinq directions découlent de la mesure : financer la capacité propre plutôt que la substitution ; ouvrir des canaux de contribution à chaque niveau ; investir dans le tressage plutôt que dans des brins isolés ; donner à l'intendance de la nature un porteur vivant ; et traiter honnêtement la question du sommet.
Integration Capacity Analysis · Discussion paper · first reading · Wim Van Laere · 30 July 2026 · The full protocol and score sheets are available on request, and published as a separate technical & replication appendix · Consistency ≠ validation.