[Kenji]

That's interesting. I once took a sociology class about the behaviour of panicking crowds and we covered exactly this phenomenon, that is, if you put a pillar in front of the emergency exit (without completely blocking it, of course), panicking people can get out faster. It seems counterintuitive but it's been shown numerous times.

[mantrax5]

It basically reduces deadlock chance.

When you have one exit and it's blocked due to many people attempting to fit at once, the entire queue is congested.

That barrier turns one exit into two exits (even if it's backed by only one real exit). If one is congested, people can still flow through the other one, and vice versa.

This explains why the evacuation rate is roughly twice faster.

If the barrier is designed to even more virtual exits, it'll be even faster.

And there's an even faster way, that requires no architectural changes, but people will have to be trained in it so well, that they'd follow it even in panic.

And this would be some heuristic which orders people in an unambiguous way, so only one person is attempting to exit at a given moment.

I don't know, left to right order, split in rows, people holding hands... Some creative solution is required to decide what that heuristic would be.

Queues have no congestion, so they'll represent the fastest possible rate of evacuation through a single exit.

[greenyoda]

"And this would be some heuristic which orders people in an unambiguous way, so only one person is attempting to exit at a given moment."

Given how badly people behave at four-way stop signs (where there's an unambiguous ordering algorithm that most drivers seem to ignore), I don't think there's much hope for a heuristic ordering working well for panicked crowds at emergency exits.

[YokoZar]

Are you sure it's unambiguous? Four way stops have both a "whoever was here first goes" rule and a "we arrived at approximately the same time so the person on the right goes" rule. That alone adds ambiguity, as they're not well defined.

But it gets worse than that, as pedestrians have priority and can break up the established order. If it's "my" turn, but a pedestrian blocks only my way, someone else can rightly cross out of turn, but then once the pedestrian is clear when does it become my turn again?

[farmfood]

I moved to the US four years ago from the UK, and I think four way stops are wonderful. I've yet to experience the disorder suggested. I grip the wheel, narrow my eyes and approach with extreme caution. I like think everyone else does too.

[Retric]

That's not really a great description of what's happening. The group in-front of the exit is moving vary slowly and everyone needs to accelerate before the exit. By limiting things to two lines people are moving faster as they approach the point of maximum contention so they don't need to accelerate as much. On top of this you add delays as people content left right or center at the exit.