Skepticism, while logically impeccable, is psychologically impossible, and there is an element of frivolous insincerity in any philosophy which pretends to accept it.
– Bertrand Russell (regarding an unreal past)
Should it be asked me whether I sincerely assent to this argument which I have been to such pains to inculcate, whether I be really one of those skeptics who hold that everything is uncertain, I should reply that neither I nor any other person was ever sincerely and constantly of that opinion. I dine, I play backgammon, I converse and am merry with my friends and when after three or four hours of amusement I would return to these speculations, they appear so cold and strange and ridiculous that I cannot find in my heart to enter into them any further. Thus the skeptic still continues to reason and believe, though he asserts he cannot defend his reason by reason.
– David Hume (problem of induction)
There is no way to distinguish between the scenarios by collecting new data.
What we’re left with is our choice of prior credences. We’re allowed to pick priors however we want—and every possibility should get some nonzero number. But it’s okay to set our prior credence in radically skeptical scenarios at very low values, and attach higher prior credence to the straightforwardly realistic possibilities.
Radical skepticism is less useful to us; it gives us no way to go through life. All of our purported knowledge, and all of our goals and aspirations, might very well be tricks being played on us. But what then? We cannot actually act on such a belief, since any act we might think is reasonable would have been suggested to us by that annoying demon. Whereas, if we take the world roughly at face value, we have a way of moving forward. There are things we want to do, questions we want to answer, and strategies for making them happen. We have every right to give high credence to views of the world that are productive and fruitful, in preference to those that would leave us paralyzed with ennui.
– (Physicist) Sean Carroll
Let’s break this down a bit.
Imagine you are trying to determine whether an unseen coin is a penny or a quarter. If you are told that it was flipped and came up heads, this shouldn’t cause you to favor either answer. Even if you were told the results of one thousand tosses, you still have learned nothing. You would need some evidence that could differentiate the cases (such as the weight of the coin).
Similarly, a “picture from yesterday” might be evidence of a real past, or it may have been planted there just now. Since it fits perfectly into either scenario, it cannot distinguish them. Therefore, as Carroll says, “there is no way to distinguish between the scenarios by collecting new data.” No evidence can tell a real past from a fake one.
Suppose you strongly believe that the coin is a quarter. If you didn’t form this belief from evidence, where did this bias come from? It can only come from your “prior credence” (a technical term in Bayesian statistics). In other words, if you can’t rely on evidence, you should feel free to say “well I just assume it is a quarter, because otherwise I’d be paralyzed with ennui!” or “a real past is more likely because WTF.”
Notice that once you’ve done this, every bit of evidence seems to justify your assumption. “Oh, it came up heads? That’s exactly what a quarter would do! It came up heads a thousand times? Now we can be totally sure!” Similarly, “this picture is obviously evidence of a real past.” It’s not, but it’s very easy to overlook that fact.