Imagine you wake up tomorrow and find yourself alone in a different universe. As a good scientist, you would like to know how that universe functions, so how do you come up with a good explanation of your universe? The most obvious requirement is that your explanation fits what you observe. If it doesn’t fit, it is at best incomplete and at worst completely false. But first you need some systematic way of checking whether a given explanation fits with your observations. Any explanation that would require you to go through some infinitely long series of steps is therefore disqualified. No such explanation would give you a precise set of data to check against the observations. So the only viable candidates are those with some endpoint.

If your lifespan in this universe is infinite, any finite explanation is as good a starting point as any other. However, you don’t know how long your life will be in this new universe, so prior to proving otherwise, you may as well assume it is finite. If the first explanation you check doesn’t fit the observations, you would like to have time to check some others before you die. Therefore, the best explanations to test first are those that have an endpoint in the very near future. That way, you’ll have time to check as many as possible in as short a time as possible. If you turn out to be wrong about your life span, you won’t have lost anything by wasting time testing overly simple false explanations.

It probably won’t take you long to realize that one of the simplest explanations, that which simply attributes every part of the universe to the actions of some God, also happens to fit all the observations perfectly. We’ll call this a radical theological view, and characterize it by saying that its God is arbitrary and operates by no predetermined set of rules. At first glance, it may seem that you have found the best possible explanation. However, after a bit more thought, you’ll realize that not only does it fit all the actual observations you make, it would also fit any other arbitrary set of observations you could possibly make. In other words, this explanation is a tautology. You cannot combine this radical theological explanation with the observations of a particular instant and make any predictions about what you would observe in another instant. This means that the explanation doesn’t really provide you with any useful information. You can simply observe one instant of the universe without any explanation at all. What you want is a way to know something about other instants of the same universe.

However, the fact that you would like an explanation with some predictive power does not mean any such explanation exists. It is possible that the universe in which you find yourself is completely random and chaotic (for instance, if the universe is run by an arbitrary God). If that is the case, then the universe is truly unpredictable and no predictive explanation is possible. Even if the universe appears to be nonrandom, it is possible that any predictive explanation turns out to be false. Any sufficiently long random sequence of elements (in this case each element is a particular instant of the universe) will contain shorter sequences that appear to be determined by some very simple set of rules. (For example, the sequence “123456” appears twelve times in the first ten million digits of pi, even though the full sequence of ten million digits has so far passed all known tests for randomness.) If you enter a totally chaotic universe at the beginning of an apparently nonrandom interval of time, you might justifiably decide upon some predictive explanation that perfectly corresponds to every observation you make in your lifetime. This explanation, however, would be wrong. It may have turned out to be very useful to you in your lifetime, but if you had lived longer, it would have become completely useless.

Despite these problems, you should still prefer a predictive explanation over one that doesn’t provide any predictions. If your universe is deterministic, a good predictive explanation can give you a great deal of information about the universe at times outside the period in which you observed things. If your universe is completely indeterministic, a predictive explanation doesn’t give you any necessarily true information, but it also doesn’t give you any less information than an explanation that can’t make predictions. You wouldn’t have lost anything by making the false assumption, because no other possible explanation would have been any more useful than the one you decided on, predictive or otherwise.