Gerald Ajam

The robot that became a social contract

The first idea was a robot.

Status Practice-derived

Published September 2026

All notes

TL;DR

We set out to automate the rearranging of classrooms. Testing showed the hard problem was not moving furniture but deciding, together, how the room is used.

The distinction it turns on

The usual reading

A logistics problem

Rooms stay in rows because moving furniture costs lesson time.

The argument here

A social problem

Rooms stay in rows because everyone knows where the front is.

01The situation

The configurator answered a real problem

In 2019 we proposed a classroom that rearranged itself. A teacher would pick a preset layout. Robotics would move the furniture. The system would track how the lesson departed from the plan, learn from it, and suggest better layouts next time.

It sounds like science fiction. The problem behind it was not. Teachers avoided active learning because rearranging the room cost lesson time, and the cost fell at the worst moment: the start of a lesson, with thirty people waiting. Remove the cost, we reasoned, and the practice would follow.

Before building anything, we tested the assumption. We set up a room with four screens and several layouts, and ran three lessons in it: Secondary 1 Science, Social Studies and English.

02The tension

In practice

The students turned round anyway

The layouts worked as layouts. Students at the back of a science demonstration watched it on a screen rather than stand up to see. Groups gathered and talked. So far, so configurable.

Then the teacher spoke. Students facing away from her turned round to face her, even with a screen directly in front of them. Asked why, they said it was a sign of respect.

The teachers moved too. Each started at centre stage and drifted to one side, to wherever they could see every student. One said she was unsure where to stand. Another disliked students facing away, because she could not tell whether they were paying attention. A student noticed the new geometry before we did: “We’re talking behind your back.” Others used the tablets to AirDrop each other funny photos across the room.

None of this was about furniture. The room had four screens and no agreed front, so everyone went looking for one.

The front of the room was never a wall. It was a social norm, and the furniture could not move it.

While students work

Each group works from its own screen. Four screens, four fronts.

When the teacher speaks

Everyone turns to her, including students with a screen directly ahead. One front again.

Where is the front? A schematic of the four-screen room from our 2019 user tests. Screens are the bars on the walls; the ringed dot is the teacher, standing to one side where she could see everyone.

03The move

From moving furniture to agreeing how the room is used

In 2020 the idea became Setup to Learn. The robot went. Two things replaced it.

  1. Give the room flexible assets, not a machine

    Instead of one system that moved everything, a set of eighteen to twenty varied items: tables that fold, stack or take writing, and pieces that make a soft boundary in seconds. The point was not automation. It was choice.

  2. Ask people to design the room they want

    In quick-learn sessions, teachers and students shaped a classroom outline with cards, each card a piece of furniture with its properties, and talked through why.

    The teachers preferred structure. The students wanted varied configurations. And some students, given a choice of everything, chose identical furniture for the whole class. Nobody could then argue about who got the good chair. Uniformity prevents disputes. That is a design insight, not a failure of imagination.

  3. Write the rules together

    The second half of Setup to Learn was a protocol for a class social contract. Students help draw up the rules, including rules for the teacher. Consequences are tied to what a student can learn from them, not to punishment, and the teacher faces consequences too.

    A flexible room raises questions a fixed one never asks. Who moves the tables? Who decides the layout? What happens when two groups want the same corner? A robot cannot answer these. A class can, if someone asks it.

04Where it breaks

What we did not show

These were three lessons, in one room, in one school. They are observations, not findings. We did not measure whether the social contract changed learning, and I would not claim it did.

Nor was the robot a foolish idea. Switching costs are real, and a room that is expensive to change will not be changed. The tests did not show that the cost did not matter. They showed it was not the only cost, and not the largest.

05Where it lands

Automating the furniture would have automated the easy part

Orientation, respect, attention and agency are negotiated, not configured. The room we tested could move in seconds and still had a front, because the people in it put one there.

I think of this often when a school asks for new furniture. The furniture is the visible part of the problem. The invisible part is the agreement about how the room is used, and it is usually the part nobody has designed.

The point

What transfers

The reason this page exists, rather than the story that produced it.

  1. The expensive-looking problem is often a social one. Check before you buy the machine.
  2. Test the assumption behind a technology before building it. Three lessons told us more than a year of specification would have.
  3. Norms are part of the design, not an afterthought to it. If a room is meant to change, someone has to agree how.
  4. Watch where the teacher stands. It tells you where the room’s front really is.
Singapore