[jancsika]

I'm not sure I understand.

Humans aren't like weighted dice. For example, it's certainly conceivable that a few pairs of these humans misunderstand the goal such that they have a 0% chance of ever choosing numbers that change the respective ordering. Add to that the higher probability that many of these pairs of humans will just toggle their orders each round.

Edit: clarification-- the humans don't even have to misunderstand the rule, just the upshot of the process.

[mongol]

Maybe you don't even need to ask them to think of a number. The person whose name is sorted before the other person's name is person 1. And they compare their lengths instead of thinking of a number?

[pdpi]

That gives you a single bit of entropy from those two people instead of a steady stream, though.

[hcs]

That's a good point, I guess you have to also randomly choose the humans?

[dmurray]

I don't think so, if they were all lined up in alphabetical order and you took them in pairs from that ordering, it would be just as good.

[pteraspidomorph]

But then you'll always get the same output from the same set of humans. It's a PRNG where the set of humans is the seed.