Artifacts
Public artifacts such as pages, reports, projects, and news items that can be traced back to their source.
Dutch Observatory
Understanding public signals around AI, digitalisation, and education.
Dutch Observatory collects indexed sources, recognises themes, and connects evidence cautiously. Its outputs are not rankings or judgements, but a tool for following developments in education.
The corpus is still growing. Read everything as an exploratory map of available sources, not as a complete view of the field.
What do you see here?
A cautious observatory, not a scoreboard
Themes and anchors
Controlled concepts connect Dutch and English variants, for example toetsing and assessment.
Organisations
Organisations are shown through observed source relations and semantic profiles, not as a ranking.
Uncertainty
New, thin, or overlapping evidence remains visible as uncertain and is not presented as a hard conclusion.
How DEO came about
An exploratory side project as observatory
DEO began as an exploratory side project to find structure in the Dutch EdTech landscape. There was no complete database, only scattered sources, reports, and organisation websites. The observatory was built to follow those sources systematically without manually judging what is important.
What was manually seeded?
At the start, a limited amount of information was entered manually: a small set of organisations, domain anchor definitions, a few dozen source feed URLs, and a small number of initiative seeds. This forms the skeleton of the corpus.
What was algorithmically discovered?
Most of the content is built algorithmically. Most relationships, semantic overlaps, anchor matches, organisation profiles, text citations, and collaboration relations come from crawling and deterministic analysis, not manual entry.
No LLM or generative AI. All relationships are computed deterministically from source fragments and are reproducible.
Technical outline Technical operation
Data sources
Configured source feeds, HTML pages, news archives, and documents. Each source has a trust level that affects artifact quality and relation evidence.
Evidence types
Text extraction, domain-anchor matching through controlled vocabulary, relation evidence through explicit wording, geography through text matching, and citations through word-boundary-aware matching.
Scoring
Quality scores based on text length, boilerplate detection, and extraction depth. No embeddings or machine learning. All rules are deterministic and reproducible.
Confidence interpretation
Publication confidence dampens public interpretation when the evidence base is thin, young, or overlapping. Uncertainty remains visible rather than hidden behind a confident-looking score.
Quick start
Choose an entry point
227 indexed
View artifacts
See what content was found, the quality of extraction, and which domain anchors were touched.
39 Organisations
Explore organisations
View observed profiles and cautious archetype suggestions based on indexed evidence.
Possible traction
Seeking Ground
Topics that may be seeking traction, but do not yet have enough evidence to read as an established pattern.
Transparency
Methodology
Read how extraction, quality, domain anchors, evidence, and publication confidence work.
Recently indexed
Traceable sources from the current corpus
Kennis delen Netwerken Hulp & Advies Digitale autonomie Position paper: visie op Big Tech en Edtech Actueel
Position paper: visie op Big Tech en Edtech
Onze visie op Big Tech en Edtech
Binnen schoolorganisaties worden steeds meer gegevens digitaal uitgewisseld en opgeslagen, bijvoorbeeld in leerlingvolgsystemen, digitale leermiddelen of via sociale media. De VO-raad vindt het belangrijk dat scholen en schoolbesturen de privacy van leerlin...
Do not infer
No judgement about importance or influence
An organisation or theme appearing more often is not automatically more important. It only means the current corpus contains more indexed evidence. Dutch Observatory supports exploration, not ranking.