how we measure.

the government's own scale.

We use the De Jong Gierveld 6-item loneliness scale: three questions on emotional loneliness, three on social loneliness. It was developed at VU Amsterdam, has been validated over decades, and is the standard instrument in Dutch public-health monitoring - the same scale behind the GGD health monitors and national loneliness statistics.

That choice is strategic as much as scientific: when we report a change, a municipal health official can place our number directly next to their own targets. A homemade well-being quiz would be easier to flatter ourselves with. That is exactly why we don't use one.

at joining, at three months, at six.

Participants are invited to answer the six questions when they join Gader - it takes about a minute - and again three and six months later. Answering is always optional and never blocks anyone from joining a gathering. Alongside the scale we record the behavioural signals that don't rely on self-report: whether someone attends again within thirty days, and whether they become part of a recurring circle.

dose and response, not before and after.

A simple before-and-after number flatters any intervention: people who feel better also go out more. So our primary comparison is dose-response: do participants who attended many gatherings change differently from those who joined once and stopped? If frequent attenders improve while low-engagement users don't, that pattern says something a raw average never can.

Where a pilot allows it, we add a waiting-group comparison - people who signed up but whose district hasn't been activated yet. And we report attrition honestly: who stopped answering, and how that may bias the result.

districts, never individuals.

Municipalities, housing corporations, and venue partners receive quarterly reports at district level: participation, repeat-encounter rate, venue activity, and measured loneliness change. Always aggregated, always with minimum group sizes, never traceable to a person. Individual answers are never shared with anyone - not with partners, not with hosts, not with other participants.

Loneliness scores are sensitive personal data and we treat them that way: explicit consent at collection, data minimisation, the right to have your answers deleted at any time, and a plain-language explanation of all of this inside the app - not buried in a policy.

what we will never claim.

We say "associated with," not "caused by." We report limits as prominently as results. We share the measurement protocol with partners before a pilot starts, so the bar is set before the data exists. And we would rather publish a modest, real effect than an impressive, unverifiable one - because one overclaimed number would cost us the only thing that makes this model work: being believed.

We don't just create connection - we measure it, with the same instrument the government uses.