There’s an island on which there lives exactly two types of people: truthers and liars. Truthers always say true statements, and liars always say false statements. One day a brilliant logician comes to visit the island. The logician knows all of the above-stated facts about the island. It also happens that the logician is a perfectly sound reasoner – he never proves anything that is false.

The logician encounters an individual named ‘Jal’ that lives on the island. The logician knows that Jal lives on the island, and so is either a truther or a liar. Now, Jal makes a statement from which it logically follows that Jal is a truther. But the logician could never possibly prove that Jal is a truther! (Remember, we never asserted that the logician proves all true things, just that the logician proves only true things). What type of statement could accomplish this?

This puzzle is from a paper by Raymond Smullyan on mathematical logic. Try to answer it for yourself before reading on!

(…)

Alright, so here’s one possible answer. Jal could say to the logician: “You will never prove that I am a truther.” I claim that this sentence logically entails that Jal is a truther, and yet the logician cannot possibly prove it.

First of all, why does it entail that Jal is a truther? Let’s prove it by contradiction. Suppose that Jal is not a truther. Then, since Jal is either a truther or a liar, Jal must be a liar. That means that every statement Jal makes must be false. So in particular, Jal’s statement that “you will never prove that I am a truther” must be false. This entails that the logician must eventually prove that Jal is a truther. But we assumed that Jal isn’t a truther! So the logician must eventually prove a falsehood. But remember, we assumed that our logician’s proofs were always sound, so that he will never prove a falsehood. So we have a contradiction.

Therefore, Jal is a truther.

Now, why can the logician not prove that Jal is a truther? This can be seen very straightforwardly: we just proved that Jal is a truther, which means that all of Jal’s statements must be true. So in particular, Jal’s statement that “you will never prove that I am a truther” must be true. So in other words, it’s true that the logician will never prove that Jal is a truther!

So there you have it, a statement that appears to satisfy both of the criteria!

But now the next question I have for you is a bit trickier. It appears from the line of reasoning above that we have just proven that Jal is a truther. So why couldn’t the logician just run through that exact same line of reasoning? It appears to be perfectly valid, and to use nothing more advanced than basic predicate logic.

But if the logician does go through that line of reasoning, then he will conclude that Jal is a truther, which will make Jal’s statement false, which is a contradiction! So we’ve gone from something which was maybe just unintuitive to an actual paradox. Can you see how to resolve this paradox? (Again, see if you can figure it out yourself before reading on!)

(…)

Okay, so here’s the resolution. If we say that the logician can go through the same line of reasoning as us, then we reach a contradiction (that a truther tells a false statement). So we must deny that the logician can go through the same line of reasoning as us. But why not? As I said above, the reasoning is nothing more complicated than basic predicate logic. So it’s not that we’re using some magical supernatural rules of inference that no mortal logician could get his hands on. It must be that one of the assumptions we used in the argument is an assumption that the logician cannot use.

So look back through the argument, and carefully consider each of the assumptions we used:

First of all, why does it entail that Jal is a truther? Let’s prove it by contradiction. Suppose that Jal is not a truther. Then, since Jal is either a truther or a liar, Jal must be a liar. That means that every statement Jal makes must be false. So in particular, Jal’s statement that “you will never prove that I am a truther” must be false. This entails that the logician must eventually prove that Jal is a truther. But we assumed that Jal isn’t a truther! So the logician must eventually prove a falsehood. But remember, we assumed that our logician’s proofs were always sound, so that he will never prove a falsehood. So we have a contradiction.

In order, we made use of the assumptions that (1) Jal is either a truther or a liar, (2) every statement made by a liar is false, and (3) the logician is a sound reasoner.

I told you at the beginning that facts (1) through (2) are all known to the logician, but I did not say the same of (3)! The logician can only run through this argument if he knows that he is a sound reasoner (that he only proves true things). And this is the problem assumption, which must be rejected.

It’s not that no logician can actually ever be sound (a logician who only ever reasons in first order logic and nothing more fancy would be sound). It’s that the logician, though he really is sound, cannot know himself to be sound. In other words, no sound system can prove its own soundness!

This is very similar to Gödel’s second incompleteness theorem. The only proof system which can assert its own consistency is an inconsistent proof system, and the only type of logician that can prove his own soundness will end up being unsound. Here’s the argument that the logician might make if they believe in their own soundness:

Supposing Jal is a liar, then his statement is false, so I could eventually prove that he is a truther. But then I’d have proven something false, which I know I can never do, so Jal must not be a liar. So he must be a truther. 

Since the logician has now produced a proof that Jal is a truther, Jal’s statement is false. This means that Jal cannot be a truther, so the logician has proven a false statement!