How Do We Know That Science Works?

Here’s a question I’ve gotten several times since starting this blog a few days ago (and several dozen times over the last few years):1

“Scientists make a lot of noise about being ‘objective’ and using data to determine the truth, but at the end of the day, science has an untested — and untestable — core belief: that we can learn the nature of reality by actively studying the physical world. So doesn’t this mean that science is a form of faith, no different from any other system of belief?”

Short answer: yes to the first part, no to the second part.

Long answer:

Okay, so let’s try to imagine not believing that physical evidence tells us about the nature of reality. How does your day look? You wake up, and you’re hungry, so you decide to go to the kitchen and pour out some cereal into a bowl. Except that you don’t, because you’re not sure that the bowl will hold your cereal — sure, it seemed like it did yesterday, but that doesn’t necessarily mean anything. Hell, you don’t even know that your cereal is in there. For that matter, how do you know that the floor will support your weight? Or that there is a floor? In fact, you aren’t even sure that eating the cereal will make you less hungry. You make dozens and dozens of tiny decisions every minute of the day, and for the vast majority of them, you have to trust that your senses give you reasonably accurate information about the world. Of course, you’re taking that fact on faith. But if you didn’t, you’d cease to be able to do anything at all in the world. I could be contrary and say that you’re taking that fact on faith too — and that’s perfectly true. It is absolutely true that I would have been screwed today if I hadn’t trusted my senses to tell me about the world, and it’s also true that I would have been even more screwed yesterday if I hadn’t trusted my senses — but that doesn’t mean that the same thing will be true tomorrow. For all we know, we’ll wake up tomorrow in a world where we can get by without paying any attention to our eyes and ears and noses.2 I wouldn’t bet on it, though — nor would you, nor any sane person.

The ultimate thing at work here is a kind of reasoning called inductive reasoning; it’s a peculiar kind of reasoning in that we humans use it constantly, with great success, but there’s no way to “prove” that it actually works. Inductive reasoning is just the idea that the more often something happened before, the more likely it is to happen again.3 So, for example, we all think it’s extremely likely that the sun will rise tomorrow, since it rose yesterday, and the day before, and the day before that, etc, throughout all of recorded history. Similarly, every time you’ve been hungry and you’ve eaten, you’ve felt less hungry, so therefore eating will probably make you less hungry in the future as well. This is, ultimately, how we know everything that we know about the world, and we use inductive reasoning so often that we hardly ever think about it. But, of course, it doesn’t have to work: just because something happened before doesn’t mean that it’ll happen again the same way, nor that it’s more likely to happen again at all. Despite this, we humans do use induction very successfully, and quite frequently; the fact that induction succeeds even though it doesn’t have to succeed is known as the problem of induction.

So it’s absolutely true that science does rest on an untestable belief: the belief that applying inductive reasoning to our perceptions can actually give us knowledge about the world. But we shouldn’t hold that against science. It’s a basic fact of logic that you can’t draw any sort of conclusions without taking some statements for granted; logicians and mathematicians call these unproven statements axioms, and you always need a few of them, even for basic stuff like addition and multiplication.4 But doesn’t this put us back to square one? If all systems of belief have axioms at their core, doesn’t that mean that they’re all equally valid and none of them should be taken as a more legitimate way of looking at the world than any of the others?

Not at all. That great untestable belief up there, that belief in the power of induction to tell us about the world? As untestable beliefs go, it’s the best one available. It’s the weakest claim that I know of that will still allow us to stumble through this world with some hope of understanding what’s going on — and as it turns out, we can do pretty damn well with nothing other than this claim, thanks to the power of Science™.

You may think I’m being very silly by making judgements about the relative merits of untestable statements. But claiming that you can’t make this kind of judgement turns out to be a really weird claim. For example, check out this pair of untestable statements:

Statement A: Applying inductive reasoning to our perceptions gives us pretty good information about the world around us.
Statement B: There is a unicorn in my basement that hangs out there, but only when nobody is looking, and it never leaves any evidence that it’s been there.5

Here’s the thing: if you don’t think that it’s possible to make judgements on the relative merits of untestable statements, then you have to say that Statement B is just as good as Statement A — and that’s just weird. I mean, invisible unicorns? Really? I guess if that’s your thing, sure, but I’m going to go out on a limb here and say the burden of proof is on you; it certainly seems like Statement A is much more plausible than Statement B, even though neither one can be justified.

So we’re back to the short answer: yes, absolutely, science is based upon an untestable belief, just like everything else! But it’s got the best untestable belief — one that you already believe, one that you could hardly afford not to believe. And that’s really the only one that we need in order to do science, whereas other systems of belief seem to require a lot of bells and whistles in addition to a belief in the power of induction. In that sense, science is not based on faith, but instead on a lack of faith relative to other systems of belief — we take as little on faith as we reasonably can when we do science.

  1. This question is usually asked in an attempt to draw a certain kind of comparison between science and religion, but not always; there are plenty of other kinds of beliefs that people have which, for one reason or another, they want to compare favorably to science. []
  2. Incidentally, this is one of the most horrifying ideas that I can think of. Just take a second and imagine a world like this, and then imagine how many people would simply retreat from their senses entirely… []
  3. Induction is actually a very wide concept, and it’s notoriously difficult to define. This is one of the broader definitions available, though. []
  4. Technically, you can prove a few things without axioms, but they tend to be statements like “Either there is a bear in this room, or there is not a bear in this room.” []
  5. See Russell’s teapot, Sagan’s dragon, and the invisible pink unicorn for more of the same. []

4 thoughts on “How Do We Know That Science Works?

  1. Well stated, though I would use the term “trust” instead of “faith” to describe how we interpret qualia.

    Model the brain as a large relational database, eg: query input from eyes -> visual cortex language center memory other related objects, actions, etc. As long as everything is working correctly, the brain tends to trust the experiences it has accrued. In that sense it’s a bit more scientific than faith: the brain is a log of experience which can formulate and test hypotheses, compare results, and remember bad assumptions all at the relatively simple scale of dendritic competition. I’d say “faith” if we were loaded with a set of knowledge at birth and weren’t predisposed to question or alter it.

    Congrats on the new blog! We always need more science blogs.

  2. Well stated.
    I realize the word “science” has more than one definition, but in general I think getting people to appreciate that science is a methodology, a process, not a belief system helps make clear the distinction.
    Echoing Z. Kemp, congratulations on the blog.

  3. I like the point about science taking on a minimal set of assumptions.

    Quibble-y question: why say that you don’t need axioms to prove instances of excluded middle?

  4. Thanks for the feedback and well-wishes!

    Zach: “Trust” is certainly a better word than “faith.” I’m not sure that you get anything extra by modeling the brain as a relational database, though. We are certainly given pre-loaded instructions to trust our senses most of the time, which is ultimately what I’m talking about.

    Bob: Yeah, the title of the post is a bit of a fudge — it should really be something like “how do we know empiricism works?” But I’m always asked this question in the context of science. And it’s certainly true that too many people see science as a body of knowledge rather than a a process for obtaining knowledge.

    K: Depending on your rules of inference, you don’t need axioms to prove instances of EM. You can, for example, do it with the standard versions of negation introduction and disjunction introduction. You can turn around and say that rules of inference are like axioms, but I’m not really sympathetic to that argument unless the rules of inference are super-weird, and I don’t think that NI and DI are particularly strange. I suppose I could have picked something even less controversial that can be proven without axioms, like “if there’s a bear in this room, then there isn’t not a bear in this room.” EM was just the first thing that came to mind.

Leave a Reply