The law as an information system
An information-science mapping of legal analysis, FLINT and DMN. A search, not a claim.
We are information specialists, not lawyers. We came to legal analysis from information management, and we build on existing Dutch work: the wetsanalyse method for legal analysis, FLINT and DMN. Where we deviate from them, we say so — so it can be judged.
One question per article: who does what?
For each paragraph of law we ask one question and classify the answer.
| Question | Class | Role in the decision |
|---|---|---|
| Does the person bring a fact? | FACT | Input — something to prove |
| Does the law exclude? | EXCLUSION | Gate |
| Does the law test a criterion? | NORM | Test |
| Does the authority act? | ACT | Output — grant or refuse |
| Does the person have a duty? | DUTY | Obligation |
Each classified paragraph also gets three dimensions:
- What does it do — its place in the decision tree;
- Who is it about — the legal subject;
- Where does the value come from — the data element and its source register.
Who and where are exactly what evidence exchange needs: which person, which fact, which authentic source.
We report what did not go as planned. When we counted, two of the five classes — FACT and DUTY — did not appear at all in the laws we processed, and three classes we had not planned for did, definitions most of all. We do not yet know why. We would like to discuss it.
Human judgement, modelled explicitly
Legal analysis stops at classification. FLINT models duties, but not the room an authority has to deviate from them. DMN is deterministic — it knows no "maybe". We add one layer: a zone per decision that marks where a person must look, for example where a hardship clause applies.
Four gaps in machine-readable law
Dutch national law is available as structured XML. But four things are missing, in every law we have looked at so far.
No link between article and service
One law can produce 23 public services. Nowhere in the structure does it say which article belongs to which service.
No label for the type of article
An exclusion, a norm test and a hardship clause look the same in the XML. The same exclusion grounds return almost word for word across laws: "rechtens zijn vrijheid is ontnomen" (lawfully deprived of liberty) appears identically in 61 articles across 32 laws, in our corpus of 319 laws and 49,629 articles.
Municipal regulations miss their legal basis
A municipal regulation does not refer to the national article it rests on in a machine-readable way. Sometimes the reference is in the running text; never in the structure.
No matrix of persons involved
The same person is a "co-resident" in one article and a "cost sharer" in the next. Without a stable identity per party, you cannot say who must provide which fact.
None of these gaps requires a change in the law. The text can stay as it is. What is missing is metadata — an index to the manual.
Why this matters in Europe
Gaps 1 and 4 are the questions evidence mapping asks: which procedure, which requirement, which person. Gap 2 means the same requirement is written down again and again, instead of once and reused.
We are exploring how Dutch legal identifiers (JuriConnect) map onto the European Legislation Identifier (ELI), so that a requirement can point to its legal basis in a form other Member States can read.
Open findings
Our findings on legal metadata are open. We share them with researchers and standardisation groups working on evidence mapping and attestation standards. Tell us you are interested and we will send them.