"" For example, consider one of the parametrizations we study: a group of 15 voters of which 3 are experts; the experts’ information is correct with probabilit

"" For example, consider one of the parametrizations we study: a group of 15 voters of which 3 are experts; the experts’ information is correct with probability 70%, while the precision of non-experts’ signals can take any value between 50 and 70%, with equal probability. Then only non-experts with signals of precision close to random should delegate: a non-expert with information that she knows is only 55% likely to be correct should not delegate to experts whom she knows to be correct with probability 70%. "" This is not about politics, or subjective questions about complicated situations. It imagines a world in which there is one correct answer and the reason that different people vote differently is due to a lack of information about that truth. This is very much a simplified "game" they're playing with. In this context, the result -- I think -- is more like an outcome of the central limit theorem. Basically, if you get lots of people each with a 55% chance of being correct to vote on a binary choice, they're very likely to pick the correct choice in aggregate. If you have one person who is 70% correct vote, they're 30% likely to get it wrong. So you actually don't want too many of the 55% people to delegate their decision to the 70% person, which is a bit counter-intuitive. This is sort of interesting, and I think says something useful about the Liquid Democracy concept. However I think the experiment is simplified to the point that it's a categorically different thing than even moderately interesting political questions, and so I don't think it's an argument against applying liquid democracy to politics.