Human in the loop

What HITL Looks Like in a Team Meeting

By Michael Polzin. Published 2026-07-01. Cluster: human-in-the-loop. Evidence classes: A, B, E, C, F, U.

Human in the loop is usually described in workshop language. Cohorts, facilitators, activities. That framing hides where most of the practice actually happens: the standing weekly team meeting, half an hour, on calendars nobody chose. This post takes the same pattern and sets it down at that table.

Same posture as the classroom write-up in Human in the loop, what it actually means: trauma-informed and non-clinical Class U. We do not claim these moves change anyone's inner state. We claim they change the number of moments where a teammate is carried past their own no by meeting momentum.

The setting nobody redesigns

Meetings inherit their shape from calendar defaults, not from the work. Thirty minutes, round-the-room updates, one loud voice, a decision that gets made in the last two minutes and revisited on Slack that afternoon. The trauma-informed care literature has been naming the same structural principles that fix this for a decade Class E: safety, trustworthiness, peer support, collaboration, voice and choice, cultural responsiveness (see SAMHSA's Concept of Trauma and Guidance for a Trauma-Informed Approach). Those are not welcome-slide words. They are agenda choices.

Three moves that fit inside a normal meeting

1. Name the opt-in on the agenda, not after the meeting starts

Before each agenda item write, in plain text, what full participation is, what partial looks like, and what stepping out looks like. All three sit on the invite. That way the choice is not a performance the quiet person has to make in front of the loud person Class C.

"10:05, blameless retro on last week's outage. Full: you share what you saw and what you would change. Partial: you listen, add one line in the shared doc if you want, no verbal share expected. Step out: you skip this item, rejoin at 10:20. None of the three costs you anything with me."

Boring on purpose. If the opt-in language is dramatic, stepping out becomes an event, and nobody uses it Class B.

2. Hold the door open during the item, not just at the top

The chat is the back-of-room chair. A short, agreed signal ("dropping to partial") in the channel is a door. The person running the meeting is not the person watching the door: assign a second, so nobody has to choose between facilitating and noticing. The private check-in still applies: one message, once, "want anything from me, or would you rather I leave you alone?" Both answers are fine Class B.

3. Let people rejoin without a spotlight

When someone comes back on video, do not rewind the meeting to welcome them. Do not pretend they were never gone. A one-line catch-up in chat, the next opt-in call resetting the whole room to a common line, and the meeting moves on. The rejoin cost stays low, so the step-out cost stays low, so the door stays open next week Class C.

What this pattern is not

Falsifiers you can run at your own table

A meeting pattern that cannot fail is not a pattern, it is a mood Class F:

  1. Count partial and step-out signals over four weeks. Zero is not a win, it is a signal the door is not really open, or the social cost has already been priced in.
  2. Ask one anonymous question at the end of the month: "Was there a moment this month you wanted to opt down and could not?" A working pattern moves that number, visibly, over successive months Class A. If it does not, the pattern is decorative.
  3. Watch decisions. If the decisions made in these meetings still get quietly re-litigated on Slack afterwards at the old rate, the meeting is producing consent by pressure, not by choice. That is on the design, not the team.

IamHITL sits inside the family program on the attainable path toward General Natural Intelligence, natural not artificial. This is a small, testable piece of that program, sized for the meeting on your calendar tomorrow morning. Run it, count the signals, ask the anonymous question, and tell us where it broke.