6. An error of logic

Given that the person who wakes up on Mars wouldn’t know whether he was real or copy, and thus whether the machine had worked or failed, why may we not infer that the original subject, should he awaken on Mars, wouldn’t know either? After all, wouldn’t the original subject, in such a case, be the person who wakes up on Mars?

The short answer is that whether or not a certain property (e.g., that one knows something) is attributable to a given individual may depend on how that individual is characterized. What this means may be explained with a neutral example.

Suppose that you are lost in a strange land, and chance upon a native whom you can ask for directions. This native is either a (truthful) knight or a (lying) knave, with equal probability, say. The question is whether you can rely on his directions. Would he be a reliable guide?

In one sense, the answer is clearly no, since his directions are equally likely to be right or wrong, and thus no better than a guess. But suppose that he happens also to be the noblest knight in the land. Then, in another sense, the answer is obviously yes, since a knight, let alone the noblest one, always speaks the truth. So one and the same individual is unreliable if characterized as the native that you chanced upon, but reliable if characterized as the noblest knight in the land.

This can happen because reliability is a “modal property” whose attribution to a given individual depends not merely on how things are with that individual, but also with how they would be with him in closely related situations, e.g., counterfactual or countertemporal ones. The native that you chanced upon happens to be the noblest knight in the land, but not necessarily so. So whether or not he speaks truthfully in various relevant counterfactual situations—something on which his reliability turns—will depend on which characterization is considered essential for the purposes at hand, and used thereby to determine which counterfactual situations are the relevant ones to consider.

Likewise, the person who wakes up on Mars is correctly deemed not to know whether he was real or copy, if we characterize him as we just did, as the person who wakes up on Mars. But the same might not be true if we characterize him as the original subject instead, i.e., in a case where the person who wakes up on Mars happens to be the original subject. That is to say, the original subject, characterized as such, may yet know whether he was real or copy. This question would be open, and would depend on various further details of the case, as I will spell out below. For accuracy, when required, I will put this by saying that the original subject qua original subject may know whether he was real or copy, even if the original subject qua person waking up on Mars does not know. For simplicity, however, when the context is clear, I will just say that the original subject may know, even if the person waking up on Mars does not know.

The point may also be put in this way. Our antagonist would like to endorse the following inference:

The person who wakes up on Mars does not know whether he is real or copy.

The person who wakes up on Mars is the original subject.

Ergo: The original subject does not know whether he is real or copy.

But this inference is really no better than this one:

The native that you chanced upon cannot be relied upon for directions.

The native that you chanced upon is the noblest knight in the land.

Ergo: The noblest knight in the land cannot be relied upon for directions.

Or this one, of a slightly different “countertemporal” sort, among many other similar examples that could be multiplied indefinitely:

I always have the telephone number of my next door neighbour.

My next door neighbour is the most beautiful girl in town.

Ergo: I always have the telephone number of the most beautiful girl in town.

We can also see the point directly if we attend again to the series of tests administered by God. To see if the person who wakes up on Mars can tell whether he is real or copy, we consider, as before, the long run performance of the person who wakes up on Mars in every test in the series. But to see if the original subject (should he awaken on Mars) can tell whether he is real or copy, we consider only those tests in which the original subject wakes up on Mars. The long run performance of the person who wakes up on Mars is bound to be mediocre, as we saw, but this does not automatically impugn the long run performance of the original subject.

Indeed, in the right circumstances, the performance of the original subject may be perfect, even as that of the person who wakes up on Mars remains mediocre. This would happen if the person who wakes up on Mars, whether original or copy, always trusts his memory (for some reason) and judges himself to be the original subject. The original subject would then end up divining his identity correctly every time, whereas the person waking up on Mars would, as usual, get it right only half the time. (The copy would always get it wrong.)