Transparency and methodology
How we work
AI-assisted analysis. A declared analytical tradition. Radical transparency.
This moment
Artificial intelligence is improving at a pace that would have seemed impossible only a few years ago. Week by week, new capabilities emerge that redraw what a machine can reason about at all.
Frontier models already perform at gold-medal level on the International Mathematical Olympiad. In May 2026, an OpenAI reasoning model refuted Paul Erdős's 1946 planar unit-distance conjecture — an 80-year-old open problem in combinatorial geometry — producing a point configuration that improves on the square lattice; the result was verified by the mathematicians Tim Gowers, Noga Alon and Thomas Bloom. These are not incremental gains — they are genuine leaps in the capacity of machines to reason.
It is time to put this capacity to work making government transparent.
Budgets are complicated. Spending decisions are opaque. The labyrinth of public-policy trade-offs has long served as good cover for waste, inefficiency, and politically convenient explanations. A budget of HUF 43,781 billion is something almost no one can read end to end. No single journalist can review 42 chapters and thousands of line items inside a single, coherent economic framework. This is not an accident — the fog has its beneficiaries.
AI changes that. Not by deciding for us — it cannot, and it should not. Democracy rests on human decisions, human values and human accountability. What AI can do is what no human team could do before: read every line item, apply the same analytical framework to each of them, model the consequences of proposed changes, and return the results in a form that citizens, journalists and decision-makers can actually use.
AI does not tell us what to value. But it does tell us, with unprecedented clarity, what the consequences of our decisions are — in the short term and the long.
In a democracy, this changes everything.
Why AI?
A new scale of scrutiny
Hungary's 2026 central budget runs to thousands of line items across 42 chapters. Until recently, the kind of chapter-by-chapter, item-by-item review this requires would have demanded a team of dozens of economists working for months — far beyond the means of any independent civil-society organisation. AI changes that equation from the ground up.
The Free Society Institute runs a multi-agent AI system — a coordinated network of specialised analytical models. The system reads the raw budget data, applies the same declared analytical framework to every line item, and produces detailed analyses, structured data outputs and policy recommendations. What once took months now takes days. What was once possible only for large, state-funded institutions is now possible for a small independent team.
This is the future of policy analysis: not AI replacing human judgment, but AI enabling human analysts to operate at a scale and consistency that were previously unimaginable.
Our framework
The framework we work with — and why
Every analysis rests on a framework, declared or not. Most government budget documents embed their assumptions without naming them. We prefer to name ours: we analyse Hungary's budget through the classical liberal tradition — the Austrian economics of Mises and Hayek, the public-choice analysis of Buchanan and Tullock, the ordoliberal rule-of-law thinking of Röpke and Eucken, and the institutional economics of North and Coase.
These traditions differ in method and register, but they converge on a common set of working premises. Prices coordinate information that no central planner can possess. Politicians and bureaucrats respond to incentives. Long-run prosperity depends on institutional quality more than on any single policy variable. Concentrated benefits with dispersed costs produce persistent rent-seeking. Intervention tends to generate unintended consequences that compound over time. We treat these as well-supported working premises, not as dogma. We publish our reasoning so it can be tested.
This does not make our analysis partisan. We apply the framework consistently, regardless of which government proposed a given expenditure or which constituency benefits from it. The framework is the lens; the data is the object; the analysis follows from both — and is open to correction wherever either the lens or the data proves wrong.
“It is not from the benevolence of the butcher, the brewer, or the baker that we expect our dinner, but from their regard to their own interest.”
— Adam Smith, The Wealth of Nations
Our claims
What we claim — and what we don't
We do not claim total objectivity — every analysis turns on choices: which framework to apply, what to emphasise, how to read the data. What we do claim is consistency and transparency:
Consistency
The same analytical framework is applied to every chapter, every ministry, every line item. There are no political favourites and no protected sacred cows.
Transparency
Our methodology is documented here. Our source data is Hungary's official published budget. Our analytical pipeline is described openly. Every conclusion is traceable back to specific line items.
Limits
Not every figure has been manually verified against the primary document. We use automated extraction and AI-assisted interpretation. Rounding errors, misclassifications and the occasional misread are possible. We work to minimise these through automation and cross-checks, but we cannot guarantee zero errors.
We expect corrections. If you find an error — a wrong figure, a misclassified item, a misread policy — please tell us. We will review and correct quickly. Rigorous debate, including challenge to our numbers, is exactly what we hope to stimulate.
The political squeeze
Truth and what is politically feasible
Democratic societies balance two things: what is true, and what is politically feasible. Economic reforms that would produce large long-term gains often impose concentrated short-term costs on identifiable groups. The incentive structure of democratic politics systematically underweights diffuse future benefits and overweights immediate visible costs.
We are fully aware of this tension. We do not pretend that a rigorous budget analysis will, on its own, transform policy. What we believe is something narrower: that the boundaries of public debate shift through the persistent accumulation of understanding — through data, explanation and open argument. What is unthinkable today becomes thinkable tomorrow, and eventually inevitable.
Our renewal programme is an explicit attempt to address this: for each major class of expenditure, we give an honest account of both the short-term disruption and the long-term benefit. We do not minimise hardship. We try to explain why we believe the trajectory leads somewhere better. The demographic briefs that walk this through stakeholder by stakeholder are being regenerated against the current renewal programme and will be published in the coming weeks.
“The curious task of economics is to demonstrate to men how little they really know about what they imagine they can design.”
— F.A. Hayek, The Fatal Conceit
Open correction
Help us be better
We correct errors openly. If you find a mistake in any of our analyses — a wrong figure, a misclassified expenditure item, a factual error — we want to know.
Email us
info@szabadtarsadalom.huWe review every submission and publish corrections where warranted.
Free Society Institute
Help us go further
If you value transparent, AI-assisted policy analysis — support the work. Every contribution keeps the research independent and free to read.