The counting thing doesn’t work; it relies on “counting” actually being a hand-wavey introduction of the exact same information that the guru introduces. The only way any deduction is possible from it is if we assume that where they start “counting” is where they have a base of common knowledge, i.e. they know that if there were only X people on the island, they would know that there are X people on the island with blue eyes. But, as I pointed out to Fixed Cross above, if there were X people on the island, they wouldn’t be able to learn that there are X people on the island with blue eyes merely by counting the blue eyes they see. This is what Silhouette has been talking about for a long time about taking knowledge out of context.
Also, even if the counting thing did work, they could not magically start. There is no logically best number at which to start counting, so the islanders’ flawless logic could not lead them to a common starting point without communicating, which is explicitly excluded in the problem statement.
“Perfect logician” is a potentially ambiguous term, but fortunately the term is defined in the problem statement: “They are all perfect logicians – if a conclusion can be logically deduced, they will do it instantly.” The phrase following the dash should be read as a definition, and no other attributes should be ascribed to a perfect logician.
James, are you arguing that you can prove what the starting number is by deduction? If you think you can, please do. Otherwise, you’re making the argument that while we don’t know, a perfect logician would know. But you must agree that perfect logician is neither omniscient nor a mind reader, correct? Are we returning to a presumption that every islander wants to leave the island? Isn’t it clear that there are simply too many unknowns preventing a perfect islander from deducing a common starting number?
On a tangent, for anyone still following that accepts that solution as the solution: One thing that’s quite fascinating about resistance to this problem is the way it happens regularly around a certain threshold: people are generally willing to accept that on an island with 1, 2, or 3 people, the islanders learn something useful from the guru. But at 4, people refuse to follow the same reasoning, and they struggle to come up with reasons. It happens at the other end, too, such that 100, 99, and 98 people can’t be expected to act without the intervention of the guru, but 97 people can. But there’s a tempting just-so explanation in evolutionary psychology that I’m inclined to accept: that we actually deal with nested knowledge in social situations, but only to a very shallow depth. We might wonder what our parents know, or what our parents know that we know, but rarely will we consider what our parents know that we know that our parents know. It starts to get head-achy around that threshold because our brains aren’t built for processing it, and they aren’t built that way because it’s only in rare and contrived situations when it matters.