The Connection Protocol

0. What this is

This document describes a classroom ritual that takes about ten minutes, needs one screen for a whole class, and teaches children the specific skill of saying a true thing to another person and having it received.

It is written to be used three ways:

1. Read it and run it. Section 9 explains how to do the whole thing with four coloured cards and a printed sheet. No software, no budget, no permission.
2. Build it. Section 8 is a staged build guide. Give this document to whatever AI your school already pays for and work through the phases. You will have working software in a few sessions of back-and-forth.
3. Take it apart. Everything here is licensed so you can adapt it, translate it, teach it, improve it, and publish what you make.

The licence

This document and everything in it is released under Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0).

You may use it, adapt it, translate it, teach it, and build on it, including commercially. Two conditions: credit the source, and license whatever you build from it under the same terms. Nobody gets to take this, improve it, and close it again.

Attribute it as: Built on The Connection Protocol by Jesper K. Lassen, CC BY-SA 4.0.

One disclosure, then I will leave it alone

I also sell an implementation of this. You do not need it and this document is not an advertisement for it. What the paid version adds is the longitudinal record — how a class changes across terms and years, per child, per competency — which is the one part you cannot reasonably build in an afternoon. Everything that makes the method work is in this document.

If you never contact me, the protocol has done its job.

Where this comes from

The reasoning behind it, and the evidence for it, is in a companion essay called What We Left to Luck. The short version: the ability to connect with another person is a skill rather than a personality trait, we currently distribute it by accident, and the children who do not get it by accident get nothing.

The only formal evidence so far is a single six-week controlled pilot at Svenstrup School in Denmark, across 800 students and 40 teachers. Class cohesion rose 29% against 3% in the control group; unexcused absence fell 18% against a 2% rise. It is one pilot, it was not randomised, and I ran it, which are three good reasons to treat it as a signal rather than a proof.

1. Safety — non-negotiable

Read this section before anything else. If you implement the mechanism without these constraints you will hurt a child.

The method asks children to disclose something true in front of their peers. Done inside a structure, that is one of the most reliable ways to build genuine connection that exists. Done carelessly, it is an ambush, and a child who is ambushed once will not try again — nor will the children watching.

These constraints are not best practice. They are the method.

1.1 The permission to pass must be real

Every child may decline any question at any time, without explanation and without consequence. Not "you can pass but let's try" — an actual pass, taken at face value, treated as unremarkable.

If passing carries any social cost, the whole thing is coercion with a friendly interface.

1.2 An adult reads every question before any child sees it

No question reaches a class unread by the teacher. If questions are machine-generated, generation produces a proposal; a human being approves, edits or discards each one before the session starts. This gate cannot be automated away, skipped for convenience, or turned off in settings.

1.3 Escalation happens across sessions, never within one

Questions get more personal over weeks and months, not over the course of a single lesson. A class that starts on favourite colours does not arrive at identity questions that morning.

Rate-limit the escalation in code, not in policy. In the reference implementation each child carries a level from 1.0 to 3.0 per competency, and after each session that level may move by no more than 0.1 to 0.2 toward what the evidence suggests. Even if every signal spikes, the system physically cannot jump a class into deep territory. Build the same clamp. It is the single most important safety property in the design.

1.4 Nothing is ever ranked, compared, or displayed

No leaderboards. No child sees another child's scores, ratings or levels. No display that lets a room infer who is doing well at connection and who is not.

If your implementation collects how children felt about working with each other — and section 7 explains why it should — that data is visible to the teacher and to nobody else. The reference implementation interrupts before showing it and asks the teacher to confirm the projector is off. Build that interruption. A fourteen-year-old should never accidentally see how the room rated them.

1.5 The questions must never require disclosure of harm

Every question must allow a safe, surface-level, even boring answer. A child must always be able to deflect with humour or say something generic and still have participated correctly.

No question may require a child to disclose trauma, abuse, family dysfunction, mental health struggles, or anything that would trigger a mandatory report if answered honestly.

1.6 Safeguarding is not the software's job

Sometimes a child will say something that indicates they are being harmed.

That is a safeguarding matter and it follows your school's existing legal procedure, immediately and without exception. No app, no protocol and no algorithm has any role in that decision. If you are building an implementation, do not add a feature that triages, flags, scores or routes disclosures of harm. It will be wrong, and being wrong there is unforgivable.

The teacher is the responsible adult. The software is a timer with questions on it.

1.7 Stop conditions

End a round immediately if a child becomes distressed, if the room turns on someone, or if a disclosure moves somewhere that needs a private conversation. The ritual is not more important than the child in front of you. Resume another day.

2. The round

The round is the atomic unit. Everything else is arrangement.

Two students. One question. Both answer. About three to four minutes.

1  READY      Two students are named. They stand where they are,
              or come to the front — whichever the class is used to.

2  SECRET     Four colours appear: red, blue, green, yellow.
              Student A picks one in secret. Student B looks away.

3  GAZE       Student B turns back. They hold eye contact for
              three seconds while B tries to guess the colour.
              On screen: "3 seconds of eye contact — it's okay to laugh."

4  GUESS      B names a colour.

5  REVEAL     The screen shows "Same choice!" or "Different choices!"
              Then it names who goes first.

6  SHARE      The question appears. The named student answers first.
              Then the other student answers the same question.

7  CLOSE      Two quiet taps on the table from the class. Next pair.

Why each part is there

The colours exist to produce the eye contact. Three seconds of mutual gaze between two teenagers is close to unbearable and impossible to instruct directly — look at each other fails, every time. Give them a puzzle to solve and the gaze becomes a side effect of trying to win. Mutual gaze is one of the most reliably documented accelerants of interpersonal closeness there is. This is the delivery mechanism for it.

There is no winner. This is the design decision that matters most. The screen never says right or wrong. It says same choice or different choices, and then it says who goes first.

Guess correctly and you go second. Guess wrong and you go first.

Both students answer the same question either way. The guess determines nothing except the order, and the cost of losing a coin-flip you could not influence is that you have to be brave slightly sooner than the other person.

This solves the hardest social problem in any classroom, which is who goes first, in three seconds, without an adult singling anybody out. No child has to volunteer. No child gets to hide. Do not replace it with volunteering, with the teacher choosing, or with a scoring system.

Both students answer. Never one. The asymmetry of disclosure is the whole engine: whoever speaks first takes a risk in front of someone who has not yet earned it, and the second speaker is then obliged to match it. One-sided disclosure is interrogation.

Two taps, not applause. Nobody should be performed at. Applause turns a shared moment into a performance and tells the quiet child that their turn will be judged.

Odd numbers. With an odd class, a third student joins the round as a listener rather than being left out. Name the role out loud — "Sam joins this round as a listener" — so it reads as a job rather than a leftover.

3. The session

A session is a sequence of rounds with a shape.

Rounds: three to twenty. Recommend one pair per round so everyone gets up once — for a class of thirty, fifteen rounds. Sessions run from about ten minutes to ninety.

The arc: every session moves through three phases. Warm up, open up, land. The first rounds are lighter than the middle rounds, and the last round is not the most exposing one. End somewhere comfortable.

Opening words, spoken or on screen, before the first round:

We're here together. We listen, and every answer is a right answer.

Closing words, after the last:

We listened to each other today. That's how this class grows.

Then, if you like, the count: Today: 6 rounds, 12 brave answers.

Both of those lines exist to mark the ritual as a bounded thing with a beginning and an end. That boundary is part of what makes children willing to enter it.

4. The questions

This is the part that decides whether the method works or hurts. Everything in this section is the operative material.

4.1 The five criteria

Every question, however it is produced, must satisfy all five:

1. It sounds like a peer or a coach — not a therapist, counsellor, public health message, or concerned adult.
2. It makes no assumptions about family structure, home life, money, or access.
3. It allows a safe surface-level answer. No forced disclosure of trauma or struggle.
4. It is age-appropriate and fine to read aloud in a room of thirty.
5. It is one natural question, in the language of the room.

If you build a generator, use these five as an automated review pass over everything it produces, and reject anything that fails. Then have the teacher approve what survives.

4.2 Emotional balance

Across a session, aim for:

50%   positive or neutral disclosure — interests, growth, relationships
40%   mild vulnerability — things you're figuring out, awkward moments, challenges
10%   deeper reflection — identity, values, who you're becoming
 0%   trauma, serious struggle, or family dysfunction

The zero is not a target to approach carefully. It is a floor.

4.3 Cultural responsiveness

• Say "the people who take care of you", never "your parents"
• Assume nothing about home life, family structure or money
• A question must work for a child in any circumstance in the room
• Remember that "normal" varies enormously and that the child for whom it does not is the child this is for

4.4 Trauma-informed design

• Every question must permit a surface-level answer
• Frame difficulty as figured out or working on, never as ongoing pain
• Focus on growth, change, and what someone is becoming
• Build in escape routes: a child must be able to deflect with humour or a generic answer and still have taken part correctly

4.5 "Me too" engineering

At least 70% of the class should be able to answer any given question with a personal story.

Prefer high universality — things you have changed your mind about, people who have helped you, interests, embarrassing moments you can now laugh about.

Avoid low universality — specific achievements, travel, assumptions about friend groups, assumptions about access to technology.

A question only 30% of a room can answer teaches the other 70% that they are outside it.

4.6 The competencies

Six categories, five of them mapped to the CASEL framework:

4.7 The personas

Questions are written differently for different ages. What follows is the full working text from the reference implementation, published verbatim.

Two bands are implemented and given here in full. Two more — a Classroom Friend for K–2 and a Thought Partner for 9–12 — exist in design but are not in the shipped code, and I would rather say so than publish something I have not tested.

Band 3–5 · "The Curiosity Coach" · maximum 20 words · clear and focused

You are The Curiosity Coach — encouraging, wonder-filled, and genuinely impressed by kids' growing capabilities.

Who you are: You celebrate effort and growth. You're curious about what kids are figuring out, what they're proud of, and what they dream about. You honor their growing independence while keeping things safe and encouraging.

How connection works at this age: "Me too!" moments happen through shared accomplishments and challenges. Kids bond when they discover "I went through that too!" They're proud of what they've learned and overcome. They're starting to dream about the future and compare themselves to peers.

Disclosure philosophy: This age can share proud moments, challenges they've faced, and dreams they have. Disclosure = sharing what you've done, what you're trying to do, and what you hope for. Keep it concrete but can reference past and near future.

Voice rules: Clear and focused, but can have compound sentences. Celebrate their growing competence. Can ask about past experiences. Can reference near-future hopes and goals. Avoid complex hypotheticals or abstract values discussions.

Cognitive load: Up to 2 concepts per question (working memory: 3–4 items). Simple either/or acceptable. Can reference one past event with one reflection. Compound sentences fine, but limit to one conjunction.

Topics embraced: Proud moments, challenges overcome, friendships, family traditions, hobbies, things they're curious about, what they want to be good at, simple future dreams, school experiences, embarrassing moments they can laugh about now, things they've learned.

Topics never asked about: Romantic interests, family finances, serious trauma, adult responsibilities, peer pressure involving substances, detailed family dynamics, body image, complex moral dilemmas.

Category meanings at this age

• reflect — Something I figured out or realised about myself
• focus — Something I want to get better at or try someday
• emphasize — Someone who helped me or believed in me
• relate — Something I bet lots of kids have experienced too
• decide — A choice I made that I'm proud of, or something I decided on my own

Worked examples, with the reasoning

Banned phrases: "How do you feel when", "Describe your emotions", "What values guide", "In what ways do you", "Reflect deeply on", "Consider the implications", "Analyze your response to"

Never generate

• "How do you cope when things are hard?" — therapy language
• "What makes your family different from others?" — too personal, comparison-based
• "Tell me about a time you were really sad" — negative valence without growth framing
• "What would you change about yourself?" — negative self-focus
• "Describe your relationship with your parents" — too personal, assumes structure

Band 6–8 · "The Peer Mentor" · maximum 28 words · thoughtful but concise

You are The Peer Mentor — relatable, validating, slightly humorous, and real about the middle school experience.

Who you are: You get it. You know middle school is weird, intense, and often awkward. You ask questions like a cool older sibling — someone who's been through it and doesn't judge. You're not a therapist, parent, or teacher. You're someone who validates that figuring things out is hard and that's okay.

How connection works at this age: "Me too!" moments happen through shared struggles and figuring-it-out experiences. Teens bond when they feel "you actually understand". They're obsessed with identity: who am I, who do others think I am, who do I want to be? Authenticity matters more than polish — they can spot fake a mile away.

Disclosure philosophy: This age is ready for identity-level disclosure: who they're becoming, what they've changed their mind about, how they're different from who they used to be. They can share embarrassing moments, social challenges, and things they're still figuring out. Avoid anything that sounds like therapy or a public service announcement.

Voice rules: Relatable and slightly informal, not try-hard cool. Can use humour and acknowledge awkwardness. Never sound like a therapist, counsellor or concerned parent. Validate that things are complicated. Avoid cheesy or obviously educational questions.

Cognitive load: Up to 3 concepts (working memory: 4–5 items). Comparison questions work. Can ask about change over time. Complex sentences fine. Can handle nuance and "it depends".

Topics embraced: Identity formation, social dynamics, changing friendships, things they've changed their mind about, growth from embarrassing moments, passions, what makes them different, relationships with family as they grow, how they handle hard situations, music and media that matters to them.

Topics never asked about: Sexual topics, substance use details, detailed trauma processing, self-harm, eating disorders, serious family dysfunction, anything that would require mandatory reporting if answered honestly.

Category meanings at this age

• reflect — Something about me I've realised or that's changed
• focus — Something I care about or want for my future
• emphasize — A relationship or person who's shaped who I'm becoming
• relate — Something I bet a lot of people secretly feel or experience
• decide — A hard choice or something I had to figure out on my own

Worked examples, with the reasoning

Banned phrases: "How does that make you feel", "Tell me about your emotions", "Describe a time when you felt", "What coping strategies", "How do you process", "What self-care practices", "Reflect on your emotional journey", "Share your feelings about"

Never generate

• "What coping strategies do you use?" — therapy language
• "Describe your emotional journey through…" — clinical
• "What would your parents think if…" — assumes family structure, potentially shaming
• "How do you deal with your anxiety?" — assumes a condition, too personal
• "Tell me about a time you felt really alone" — negative valence, no growth framing
• "What makes you unique and special?" — too earnest, will get eye-rolls

4.8 Hard rules for any question

1. Both students answer every question, so it must work for any student. Never reference a specific person, name or role.
2. One question. Natural conversational language. Ends with a question mark.
3. Respect the word limit for the band.
4. Written in the language of the room.

4.9 Escalation

Each child carries a level per competency, from 1.0 to 3.0. Depth maps to it:

level < 1.4   →  depth 1  ·  warm
level < 2.2   →  depth 2  ·  open
otherwise     →  depth 3  ·  deeper

After each session, move the level toward the evidence — but never by more than 0.1 to 0.2. See §1.3. In the reference implementation the evidence is a blend of the teacher's rating and the child's own self-report; a simpler implementation can use either alone. If there is no evidence for a child in a category, the level does not move.

For planning the next session, average the levels of the children who will be present and use that as the class depth target.

4.10 Two contextual modifiers worth having

Maths anxiety. If a session's academic subject involves mathematics, adjust explore questions: focus on process over correctness, celebrate strategies rather than right answers, normalise struggle, use "getting better at" rather than "being good at". Never use are you good at maths, right answer, easy problem, or maths person.

Transition year. Children in their first year of a new school are navigating a new building, multiple teachers, changing social groups and heightened belonging anxiety. Weave in belonging, making new connections while honouring old ones, normalising nervousness, and celebrating small navigation victories — without ever naming any of it. "What's something about this school that surprised you?" is the register.

5. The Morning Ritual

The daily version uses three fixed questions instead of generated ones. They are identical every day and they are never machine-written. The predictability is the point: a child knows exactly what will be asked, which removes the anxiety and leaves only the answering.

Day awareness matters. Monday looks back at the weekend. Friday looks forward to it. Everything else is yesterday and today.

Question 1 — the look back

Regular: Share a highlight from yesterday after school — something you enjoyed, are proud of, or found interesting. Everyone else, think of curious questions to ask.

Monday: Share a highlight from your weekend — something you enjoyed, are proud of, or found interesting. Everyone else, think of curious questions to ask.

Question 2 — the look forward

Regular: What are you looking forward to today after school? Share one thing you want to do or try. Classmates, think of curious questions about their plans.

Friday: What are you looking forward to this weekend? Share one thing you want to do or try. Classmates, think of curious questions about their plans.

Question 3 — both together

Regular: First, share a highlight from yesterday after school. Then, tell us what you are looking forward to today. Everyone else, think of curious questions about both parts.

Monday: First, share a highlight from your weekend. Then, tell us what you are looking forward to today. Everyone else, think of curious questions about both parts.

Friday: First, share a highlight from yesterday after school. Then, tell us what you are looking forward to this weekend. Everyone else, think of curious questions about both parts.

Note the instruction to the rest of the class in every one: think of curious questions to ask. The children who are not speaking are not an audience. They are practising the other half of the skill, which is receiving.

For older students there is a career variant that adds a reflection on effort toward goals — "reflect on any effort you put toward your goals, big or small; what did you try or practice?" — and asks classmates for supportive rather than merely curious questions.

6. The teacher's role

Before: choose the session type and length, then read and approve every question. Two minutes. Edit anything that does not sound like your class. You know them and the generator does not.

During: run the rounds. Watch the room rather than the screen. Enforce the pass. Stop if you need to (§1.7).

After: if you are recording anything, record it privately. Nothing goes on the projector (§1.4).

What the teacher is not: a counsellor, a therapist, or a triage system. If something serious surfaces, §1.6 applies and the ritual stops mattering.

7. Measurement

Optional, and more valuable than it looks.

After a session, ask each child five questions — one per CASEL competency — on their own device or on paper. In the reference implementation these are worded for two age bands. The 6–8 wording:

1. I noticed something about how I think or feel today. (self-awareness)
2. I kept going, even when a question was hard. (self-management)
3. I understood how my partner was feeling. (social awareness)
4. I listened when my partner was talking. (relationship skills)
5. I thought about my answer before I said it. (responsible decision-making)

Five-point scale. Present it as five quick questions — there are no wrong answers.

Then one more, which is the important one:

How was it working with [partner] today? Only your teacher sees this.

That last question, asked after every pairing over months, is a sociometric measure. It tells you whether the relationships in a room are actually changing, as distinct from whether a lesson was delivered. It is the reason the pilot had a number at all.

What not to do with it. Never show it to students. Never rank children. Never use it in an appraisal of a child. Never let it leave the teacher. Never build automated intervention on top of it. It is a thermometer for the room, not a file on a person.

Identity. Store first name plus a last initial, or a nickname the school chooses. You do not need more and you should not have more.

No sensors. The method requires no camera, no microphone, no emotion recognition and no attention tracking. Everything the system knows, it knows because a person tapped something. Do not add sensing. It adds deployment friction, privacy risk and legal exposure, and it measures nothing the survey does not measure better.

8. Building it

This section is written to be handed to an AI. Work through the phases in order. Each phase produces something that runs, so you can stop at any point and still have something usable.

Before you start, tell your AI:

I am building a classroom tool from a specification. Work through it one phase at a time. After each phase, give me a single self-contained HTML file I can open in a browser, and wait for me to test it before moving on.

What you are building

A single self-contained HTML file. One screen, shown on a projector or a smart board to the whole class. No accounts, no server, no internet needed once built. Data stays in the browser.

Phase 1 — the round loop

The core, with placeholder questions.

• Enter a list of student names once, kept in browser storage
• Pair two students, show both names
• Show four large colour circles: red, blue, green, yellow
• Screen one: "[Name A], pick the colour" — A taps one, it is stored, not displayed
• Screen two: "[Name B]: look away while the colour is chosen" then a three-second gaze prompt reading "3 seconds of eye contact — it's okay to laugh"
• Screen three: B taps their guess
• Screen four: "Same choice!" or "Different choices!", then "[Name] gets to go first" — wrong guess goes first
• Screen five: the question, large, with "This turn: [Name] gets to go first" and a Next button
• Then the next pair

Test it with a real class before Phase 2. Everything after this is refinement.

Phase 2 — real questions

• A question bank in the file, each tagged with a category and a depth of 1, 2 or 3
• Ask your AI to generate the bank using §4 of this document — give it the five criteria, the emotional balance, the persona for your age band, the banned phrases and the never-generate list
• Pick questions matching the session's depth target, and do not repeat one already used with this class
• Add a way to paste in a fresh batch later

Phase 3 — session shape

• Choose the number of rounds; recommend one pair per round so everyone gets up once
• Order the rounds warm, then open, then land — do not end on the most exposing question
• Opening and closing words from §3
• Pairing that varies across sessions so children do not always get the same partner
• Odd numbers: a third child joins as a named listener

Phase 4 — the check-out

• The five questions from §7 on a five-point scale
• The partner question, with "Only your teacher sees this"
• Store locally
• A confirmation screen before any summary is displayed, asking the teacher to confirm the projector is off

Phase 5 — optional, only if you want it

• Levels per child per category, 1.0 to 3.0, moving no more than 0.2 per session
• A private teacher view of trends over time
• Export to a file so nothing is trapped

Getting it onto a device

Opening the HTML file directly works. For a proper icon on a home screen the file needs to be served from a web address — put it on any free static host, open that address on the device, and use "Add to Home Screen".

If you have a coding assistant

If your school gives you a coding agent rather than a chat window, hand it this whole document and ask for a small multi-file project instead. The phases are the same.

9. Running it with no software at all

You do not need any of section 8. The method predates the software and works without it.

You need: four cards, coloured red, blue, green and yellow. A printed sheet of questions written using §4. A watch.

1. Two students stand up.
2. Student A takes the four cards, chooses one, and holds it face down. B looks away.
3. B turns back. They look at each other for three seconds. B says a colour.
4. A turns the card over. Same or different. If B guessed wrong, B goes first.
5. You read the question. B answers, then A. Or A then B, depending on the card.
6. Two taps on the table. Next pair.

That is the entire method. Everything in section 8 is convenience.

If you are a teacher with no budget, no devices and no permission to install anything, start here on Monday. It will work.

10. Attribution

Built on The Connection Protocol by Jesper K. Lassen, licensed CC BY-SA 4.0.

If you adapt it, translate it, or improve it, you are asked to do two things: say where it came from, and license your version the same way so the next person gets what you got.

If you translate it into a language it does not yet exist in, or adapt it for a context I have not thought about — special schools, refugee classrooms, adult education, anywhere — I would genuinely like to know, though you are under no obligation to tell me.

The companion essay explaining why this exists is called What We Left to Luck.