Kindred Research
Every Circle should teach the engine something.
Kindred is building a longitudinal record of how context, group formation, guided activities, behavioural signals, interventions, handoff and later outcomes relate to one another.
Patterns become hypotheses. Replication and real outcomes determine what earns its way into the engine.
- contextcohort · defined population · consent scope
- circle6 members · formation-v1 · keep-apart honoured
- journeypinned journey version · activities logged
- signalssignals-v1.2 · observed vs self-reported
- interventionpolicy-v1.4 · decision + alternatives + reason
- handoffhandoff-v1.1 · pacing + safety gate
- outcomesday 7 / 30 / 90 · unanswered kept as null
- learninghypothesis only · awaiting real-human evidence
Every completed journey adds one structured, versioned record.
The framework
Know, test, and hold back.
What we know
- How to form Circles deterministically and explain why
- How to log every decision against a frozen version
- How to follow up at 7, 30 and 90 days without counting skips as no
- How to suppress small groups in organisation reports
What we are testing
- How much common ground a Circle needs, and how much difference helps
- The earliest point guidance can taper without harming independent contact
- Which early signals are informative about day 90
- When a bridge helps a peripheral member, and when it gets in the way
- How campus, work and events differ
What we will not claim yet
- Any causal effect of Kindred on outcomes
- Retention, wellbeing, productivity or belonging effects
- That current thresholds and weights are optimal
- That any predictive model is validated
Standards
Scientific discipline, in plain terms.
01
Versioning
Every cohort runs on frozen versions. A change means a new version, so results stay comparable.
02
Experiments where possible
Primary outcomes are defined before assignment. Small samples are labelled exploratory.
03
Correlation is not causation
Observational patterns are never reported as causal. Causal claims need a controlled design with adequate power.
04
Privacy and minimisation
Pseudonymous identifiers, explicit research consent, and only the data the question needs.
05
Human oversight
Safety cases, handoff confirmation and data requests are handled by a person.
06
Outcomes after handoff
Activity inside Kindred is telemetry. What counts is what continues after it steps away.
Every completed journey adds structure to the question: under what conditions do new relationships continue without us?
Current thresholds and weights are documented hypotheses. They will change only through new, registered versions once real human evidence supports it.
Research partners
Run one population with us, and learn from it together.
No rip-and-replace. No new community to maintain. One defined population, one measurable relationship journey.
We’re currently working with a small number of early partners.
- 1Choose an upcoming event, intake or cohort
- 2Agree primary outcomes and consent scope up front
- 3Kindred forms and guides small Circles
- 4The engine hands off when criteria are met
- 5You receive 7/30/90-day aggregate outcome reporting