Evidence-first social matching research

From first meeting to durable tie

Profile similarity can help decide who meets. The next research question is whether reciprocal experiences during and after that meeting better identify the rare connections that endure.

+4.48 ROC-AUC points beyond similarity

11.2× lift among the top 1% of ranked dyads

90–180 days held-out durable-interaction window

These results come from a longitudinal online cohort. The illustration below is synthetic and is not an IRL prediction.

A group-experience translation

Measure what happened between people—not only who looked similar

Imagine six strangers at one table. Guest A and Guest B are the selected dyad. Each person answers the same candidate items independently after the experience.

A B C D E F Selected dyad
A ↔ B

Profile questionnaire
Build the similarity-only baseline.

Directed feedback
Measure reciprocal experience signals.

Longitudinal follow-up
Observe which ties remain active.

Synthetic dyad explorer

See how two directional responses become dyadic features

Use a seven-point agreement scale. These are not validated measures and the page deliberately produces no compatibility score, probability, or relationship advice.

Emotional safety

I felt safe expressing uncertainty or disagreement with this person.

Guest A about B: 6 Guest B about A: 5

A: Not enough information B: Not enough information

Feeling understood

This person responded in a way that made me feel understood and accepted.

Guest A about B: 5 Guest B about A: 5

A: Not enough information B: Not enough information

Personal sharing

Personal sharing deepened in both directions rather than remaining one-sided.

Guest A about B: 4 Guest B about A: 6

A: Not enough information B: Not enough information

Future investment

I would make time to interact with this person again.

Guest A about B: 6 Guest B about A: 3

A: Not enough information B: Not enough information

Candidate dyadic features. Missing directional responses remain missing.
Construct	A → B	B → A	Minimum	Mean	Gap

Prospective measurement

Initial connection is not durable connection

Immediately Per-person experience
Safety, responsiveness, sharing, and desire to reconnect.
Day 7 Observed investment
Who initiated, responded, or accepted another experience.
Day 30 Repeated interaction
Contact across distinct days and any additional meeting.
Days 90–180 Durable-tie outcome
Mutually reported continued interaction in the held-out window.

The decisive experiment

Test incremental value, not a universal compatibility formula

Baseline Profile similarity

Values, preferences, goals, and pre-event compatibility.

Thesis features Interaction dynamics

Reciprocal post-event perceptions and observed follow-through.

Evaluation Combined model

Held-out lift beyond similarity, with temporal and event-group splits.

The useful deliverable is an honest number: how much reciprocal experience signals improve rare durable-tie identification beyond the existing matching baseline.

Read the IRL group-study design Inspect the reproducible benchmark