The physics of a conference room

This one's a little different — it's a model I built for fun, ported into the Lab. There's no dataset behind it; the "data" is the behaviour that emerges from a handful of simple rules.

Everyone who's been to a conference knows the pattern. People arrive, mill about, and within twenty minutes the room has sorted itself into knots — mostly people who already knew each other, or who work on the same thing. A few restless souls drift between groups. This is an attempt to capture that with physics.

The two kinds of people

Every agent has a discipline (its colour) and one of two temperaments:

• Clique-prone — drawn to others of the same discipline. Left alone, they clump into tight single-colour groups. The slider controls how much of the room is like this.
• Socially open (marked with a +) — the connectors. They seek out disciplines they haven't met, bridge groups, and — crucially — get restless. When a group grows too homogeneous, they leave.

Watch the open agents

The interesting behaviour all comes from the open agents, and the rings and arrows around them tell you what they're doing:

• A gold ring — bridging two or more disciplines at once. This is the good state: it's what mixing looks like.
• A blue dashed line — seeking a discipline it hasn't met, reaching across the room.
• A purple arc, then an arrow — a group has grown too big, so the agent splits off and drags others toward fresh territory.
• An orange arrow — the group has become a monoculture and the bridge has gone dead, so the agent escapes.
• A teal arrow — the agent is simply satisfied and leaving to find something new.

The whole model is a tug-of-war between two urges: the comfort of your own kind, and the value of everyone else's.

The metric that matters

The number to watch is cross-pollination. It rewards groups that are genuinely mixed (measured with a diversity index) and penalizes a room where all the action happens in a few isolated huddles. Slide "clique-prone" up toward 97% and watch it collapse — a room full of homophily mixes almost nothing, no matter how busy it looks. Turn it down and the open agents start stitching the room together.

Why it's here

It isn't epidemiology or economics, but it's the same instinct as everything else in this lab: take a system you can only describe in words — "conferences self-segregate" — and turn it into something you can actually watch happen, and measure. Sometimes the point of a simulation isn't to predict anything; it's to make an intuition visible.