How the work is done.
Hagsmunasamtök Æskunnar is an analysis organization. Our role is to gather what the public sector publishes about children's welfare into one place, connect it, and follow up on promises and improvements. The system that carries the work is under construction and will be rolled out in stages. This page describes how it is designed, what rules govern everything we publish, and what comes out of it.
Where do we start?
We start narrow, deliberately. The organization's first focus is the situation of children with complex needs and treatment services for children, not least for alcohol and substance-abuse problems. That is where the need is most pressing, and it is where the National Audit Office confirmed that children still fall between systems and lack services. As staffing, funding, and experience allow, we widen the lens. The system is designed so that each new area joins the same chain; nothing has to be built from scratch.
- Founded in March 2026
- On the public-benefit registry
- System under construction
- First focus: children with complex needs
The system we are building.
At the core of HÆ's work is software that gathers public data about children into one place and connects it so the context is visible.
The primary-source archive
Every document the system retrieves is kept unchanged, exactly as it arrived: with its origin, date, a direct link, and a digital fingerprint that proves it has not been altered. If a publisher changes a document, the older version is kept too, so the history stays visible. This is the foundation of our rule: every claim comes with a primary source.
Cleanup and classification
The text is extracted from every document, including scanned pages, and the key details are recorded in a fixed form: who published it, when, and about what. Promises, decisions, and figures each get their own format so they can be compared across years and institutions, and everything is classified against the same criteria, with the Convention on the Rights of the Child as the yardstick. The machine's suggestions carry a reliability score; anything it is unsure about goes straight to a person.
Profiles and connections
Every institution, every law, every service, and every promise gets its own profile, and each document links to the profiles it concerns. That turns the archive into a network: from a service's page you can trace all the decisions, allocations, and reports that concern it, and a repeated promise links back to the original. The profiles are about the system, not people. We do not record individual cases.
Monitoring and the incident timeline
The system listens to the public sector's feeds. Every new document becomes an event on the timeline of the profile it concerns: a new regulation on the timeline of a policy area, a review on an institution's timeline, an allocation on a service's timeline. A promise's status changes only when a document proves the change, and an alert flags any movement in a monitored area. That way it is plain to see what is decided, what is delivered, and what actually changes.
Human review
Nothing is published without a person having gone over it. The machine drafts entries and ranks them by reliability; a person approves, edits, or rejects, and every decision is logged. In the early stage every single entry goes through human review, and substantive judgment is not automated. Errors found after publication go into a public corrections log.
Advocacy and communication
On the basis of the analysis, we present findings to government, parliamentary committees, practitioners, and the public, submit formal consultation responses on bills and regulations, and take an active part in public debate. We keep track of the public commitments governments make and follow up firmly on whether approved projects are actually carried out.
Here is what it looks like in practice
Here is what it looks like in practice. Take an example. A minister promises from the floor of Alþingi that treatment places for children will be increased by year's end. The speech comes in through Alþingi's feed the same day, and the original goes into the source archive. The system proposes an entry: who promised what, when, and by what date; a person reads it over and approves it. The promise links to the profile of the treatment services and goes onto the timeline. It is not alone there: the same profile holds earlier allocations, oversight-body reviews, and waiting-list summaries, and the system also remembers that a similar promise was made before and how it turned out. Months later the budget bill appears, and the system flags whether the allocation is in it or not; the status updates with the document that proves it. If a regulation on the service goes to open consultation, we get an alert and submit a response before the deadline. And because everything is classified against the same criteria, the situation can be compared with neighbouring countries, year by year. Our consultation response and end-of-term review then set out plainly how it went.
Where does the data come from, exactly?
The public sector publishes more than most people realise. Alþingi maintains open data and RSS feeds on parliamentary business, committee opinions, and votes; the Government Offices and individual ministries publish news, plans, and reports; the state consultation portal opens draft bills, policy intentions, and responses before matters become law; Statistics Iceland shares its figures through an open web service (PxWeb); and opingogn.is gathers institutions' open datasets in one place. The independent oversight bodies hold a special place, drawing out what goes wrong: the National Audit Office with its administrative reviews, the Ombudsman for Children and the Parliamentary Ombudsman with opinions and observations, the Quality and Oversight Authority for Welfare monitoring treatment and child-protection services, and the National Agency for Children and Families and the Directorate of Health with statistics and reviews. Added to that is the Government's Child Well-being Dashboard. We also draw on reliable international sources such as the observations of the UN Committee on the Rights of the Child and reviews by the Council of Europe. Our system connects to these feeds directly; it is the most reliable route to the data.
What is not published, we request under the Information Act, which gives the public a right of access to government data. That right also reaches private parties to the extent that they have been entrusted with a public service, for example through a service contract, and rules on the re-use of public information allow new services to be built on the data. We send these requests systematically and keep track of how they fare.
And then there are the people. Alongside the data, we listen to front-line practitioners who know the system first-hand; that insight is invaluable, because it tells us where to look and how the system plays out in practice. But insight is context, not a source: only traceable data do we publish as fact.
This list is not exhaustive. There are more reliable public sources, and new ones are added all the time.
Five steps before anything is published.
Verification
Every document has a traceable, public origin. The origin is always recorded, and originals are never altered.
Classification
The data is classified by subject area, time period, and clear definitions, and every machine suggestion carries a reliability score.
Linking
The information is linked to related data, domestic and international, so the context is visible.
Registration
Each entry goes into the database with a reference to its primary source, a timestamp, accompanying notes, and a review history: who approved what, and when.
Assessment
Practitioners and specialists review all material before publication, and material that does not meet the requirements is not published. We analyse the system, not individual cases.
The rule is simple: we claim nothing without a source. And we welcome any feedback that makes the data better.