Monday, July 16, 2012

Why Falsificationism Sucks

Home  >  Philosophy  >  Philosophy Of Science

I once heard an atheist accuse theism of not being falsifiable. That’s a bit odd when you think about it, because “theism isn’t falsifiable” is basically a tacit admission that the problem of evil fails to refute theism. Still, presumably “theism isn’t falsifiable” is supposed to be some sort of objection against the rationality of theism, which thus inspired this article. In this entry I’ll talk about how well falsificationism works as a criterion of meaning, rationality, and scientific legitimacy.

Varieties of Falsificationism that Suck

Falsificationism comes in many varieties, but two forms I’ll talk about are:

    (1) For a claim to be meaningful, it must be falsifiable.
    (2) For a claim to be rational, it must be falsifiable.

I’ll argue that both varieties of falsificationism suck.

It’s easy to think of counterexamples to (1) and (2). “There is no greatest prime number” is meaningful and rational to believe (at least, it’s rational to believe once one apprehends the mathematical proof for it), yet it is logically impossible to falsify it and is therefore non-falsifiable. Why? Because it’s logically impossible for “There is no greatest prime number” to be false, ergo it is logically impossible to prove that it is false.

Still, one might think this can be remedied. Say that a claim is logically necessary if and only if it is logically impossible for it to be false. Say that a claim is logically contingent if and only if neither its truth nor its falsity is logically necessary. We can modify (1) and (2) to get the following:

    (1*) For a logically contingent claim to be meaningful, it must be falsifiable.
    (2*) For a logically contingent claim to be rational, it must be falsifiable.

Yet even these improved versions suck. Consider the following claim:
I exist.
That claim is logically contingent, yet it is logically impossible for me to falsify this claim, for to falsify it I would have to exist. In spite of all that, the proposition I exist is meaningful and rational to believe. One could try to remedy this by saying something like, “By a claim being falsifiable I mean that it’s possible in principle for somebody to falsify it.” In that case we have the following counterexample to (1*) and (2*):
Somebody exists.
This claim is logically contingent, yet it is logically impossible for anybody to falsify it. In spite of all that, the proposition Somebody exists is both meaningful and rational to believe. One could say, “Falsificationism applies everywhere except for the proposition Somebody exists.” That sort of falsificationism seems a bit ad hoc, but even with this version we can find a counterexample. Consider the following claim.
There is, has been, or will be a bucket of ten red balls.
We can’t even in principle falsify this claim; no matter our past experience, there is still at least the outside chance that someday in the future we will have discovered it. Not only that, but this claim can in principle be empirically verified and rational to believe.

One general reason falsificationism varieties like these suck is there are meaningful, empirically verifiable, rational claims where it’s easy to justify the truth of the claim but is trickier to prove the claim false. In any case, as a criterion of rationality and meaning, falsificationism fails. The fact (if it were so) that theism is non-falsifiable is insufficient to prevent it from being a rational belief.

Falsificationism and Science

Some varieties of falsificationism relate to science. One could say that for a belief, theory, or hypothesis to be scientific it must be falsifiable. Why this might be problematic should by now be apparent. As we’ve already seen, some non-falsifiable beliefs are empirically verifiable and rational to accept. So if science is to be a rational project in the truth business, it seems that non-falsifiability shouldn’t automatically disqualify a view from being scientific.

Consider for example the theory (or hypothesis, if one prefers) of abiogenesis: the view that undirected natural processes evolved life from non-life. Barring extreme cases like time travel and psychic ability, this view is not falsifiable, for what observation could possibility falsify it? Perhaps we could find obstacles for undirected natural processes, but any obstacles we find one could just say, “There is a way for undirected natural processes to overcome those obstacles and we just haven’t discovered it yet.” In response one could say, “That maneuver wouldn’t work if the obstacles in question were physical laws; if we found that for abiogenesis to take place it would require violating heavily verified physical laws, that would falsify abiogenesis.” That would be an interesting scenario, since in this case it seems we would (if we also knew life began to exist) have grounds for thinking that life was supernaturally created. Still, even in this case we wouldn’t have falsified abiogenesis. The abiogenesis proponent could say that perhaps physical laws as we know them today didn’t apply everywhere on earth billions of years ago.[1] While many of us take it for granted that the physical laws as we know them today applied everywhere on earth those eons ago, we haven’t actually proven that to be true (none of us were even around back then to test the laws anywhere on earth). So abiogenesis would not, strictly speaking, be proven false in this scenario even if we would have some rational grounds for rejecting it. Nor is this sort of thing limited to abiogenesis. Even if there were a physical law that forbid large-scale Darwinian evolution, the Darwinist could, similar to our hypothetical abiogenesis proponent, say that maybe the physical laws as we know them today didn’t apply back then at certain locations under unknown conditions.

It is of course possible for a scientific theory to make predictions and have those predictions be falsified, but theories do not make predictions out of a vacuum. Rather, the predictions of a theory arise from a background system of beliefs called auxiliary hypotheses or auxiliary assumptions (some of which are other scientific theories), and the truth of such auxiliary hypotheses are typically not rigorously proven. For example, it was once thought that if the Earth was really spinning on its axis, birds would get blown west whenever they let go of a tree branch. We no longer accept that as evidence that the earth is not spinning because we have accepted a different background system of physics that allows us to make different predictions. Still, we can have rational grounds for thinking that our background system of beliefs is correct, and so if Darwinian evolution did conflict with physical laws, we would have rational grounds for rejecting the theory, even though we can’t prove that exceptions to physical laws didn’t take place back then to allow Darwinian evolution to occur. In science, one cannot prove theories true in the sense that we can prove theorems of logic and mathematics; there’s always at least the outside chance that our theories are wrong (and if history is any indication, at least some of our scientific theories are in fact probably wrong in some respect); similarly, we often cannot prove theories false due to the reliance of auxiliary hypotheses. The reliance on auxiliary hypothesis to make empirical predictions (and thus the inability to isolate one hypothesis for empirical testing, since the necessity of auxiliary hypothesis means that hypotheses are tested in bundles) is called the Duhem-Quine thesis or the Duhem-Quine problem. More precisely (since the definition of the “Duhem-Quine thesis” can vary somewhat) the “theories can be empirically tested only in bundles” belief is also called confirmation holism.

We could use another form of falsificationism, perhaps something like “No theory is scientific unless it is possible in principle to have sufficiently strong evidence against it.” To use a specific example, consider intelligent design (ID) theory, which as applied to life on earth is the theory that a designing intelligence created certain features of life.[2] A popular criticism against ID is that it is not legitimately scientific and that ID is non-falsifiable. Why? Suppose we found a detailed, rigorous account for how undirected natural processes could evolve the first single-celled organism and how that could lead to the evolution of mice, reptiles, and all of life’s diversity on Earth. Does this refute ID theory? No, because even if natural processes are reasonably capable of doing the job, it’s still possible for an intelligent designer to do the same thing (albeit possibly needing technology far beyond ours). It’s hard to have real solid evidence against ID theory, and so in some sense it is non-falsifiable and therefore not legitimately scientific.

But that objection is problematic. To see why, consider the following scenario. Suppose we live in a technologically advanced society where a designing intelligence is capable of creating mice, gold (through subatomic rearrangement), electrons (via converting photons into electrons), and anything else nature could create within certain size limits. There is a box with something inside of it but neither of us knows what it is. The person who put the thing in the box has died and has destroyed all evidence of its origin, apart from the object itself. I present the theory, “What is in this box was created by a designing intelligence, as opposed to undirected natural processes like geological formation.” This theory is non-falsifiable, since anything that nature could have produced a designing intelligence could also have produced (note the similarity between this and ID). Still, if I open the box and find a desktop computer, it seems we would have rational grounds for thinking a designing intelligence created it. The fact that “What’s in this box was created by a designing intelligence” is non-falsifiable wouldn’t change that, nor would it be any reason to reject this sort of intelligent design theory.

Of course, human beings are not computers, but there do seem to be conceivable circumstances where scientists could have rational grounds for thinking ID theory is true. If for example scientists examined the physiology of organisms and discovered that all life forms on earth are actually robots (including us), it seems scientists would have good evidence for ID theory being true. One might say if we had sufficient evidence for it, ID would be science, but since it doesn’t, it is not science. Even if that’s true, the fact remains that non-falsifiability is insufficient for preventing a theory from being legitimately scientific. If it were sufficient, then it would be the case that science should not accept ID even if the scientific evidence overwhelmingly supported it, and that doesn’t seem quite right.

In response, one could bite the bullet and say, “Yes, that may be the case, but science is a human activity and we can place whatever restrictions we wish.” Maybe that’s true, but according to the falsifiability criterion, even scientifically well-supported theories we know to be true must be rejected by science, and this seems no more sensible than saying scientifically well-supported theories we know to be true must be rejected by science if the theories are created by somebody named Bob. Science is indeed a human activity and we can set up any restrictions we wish, but if we set up restrictions like “a theory that is non-falsifiable is sufficient for science to reject it, truth and evidence be hanged” and “a theory being made by a guy named Bob is sufficient for science to reject it, truth and evidence be hanged” we can no longer say that science’s goal is to rationally obtain the truth about the natural world; it becomes more like a game. If one wants to have science still be a rational project in the truth business, one could say ID should be rejected because there is insufficient evidence for it. Maybe that’s true, but then ID should be rejected for that reason, not because it isn’t falsifiable. A wrong theory should be rejected for the right reasons.


Falsificationism sucks horribly as a criterion for meaning and rationality. There are a number of meaningful, rational, empirically verifiable statements that are not falsifiable. In science, if we demand that beliefs/theories/hypotheses need to be capable of being proven false to be scientific, this doesn’t work either; both abiogenesis and evolution cannot be proven false, even though it is conceivable to have evidence against it. A better form of falsificationism with respect to science is “it is possible in principle to have sufficiently strong evidence against it,” but even this doesn’t work if we’re to have science be in the business of truth and rationality. After all, it is possible for a claim to be meaningful, rational, empirically verifiable without being falsifiable even in this weak sense as the “What’s in this box” scenario illustrates.

[1] One might think, “Would anyone really be willing to go this far?” Consider this real-life case. To support a claim I made in an Internet dialogue with an atheist, I said if things can pop into being uncaused out of nothing, it becomes inexplicable why anything and everything doesn’t pop into being uncaused out of nothing. The atheist suggested that perhaps such things did happen somewhere in the cosmos outside of our observable sphere; apparently that this would conflict with the conservation of mass-energy wasn’t enough for him to reject this as a live option. To be sure I couldn’t rigorously prove that violations of physical laws as we know them aren’t happening somewhere far out in space, even though astrophysics tends to rely on such uniformity of nature.

[2] I’m actually being somewhat uncharitable to ID here, since more falsifiable formulations of the belief are possible. If for example we go by how the most prominent ID organization describes the theory, when applied to life ID holds that certain features “of living things are best explained by an intelligent cause, not an undirected process such as natural selection.” That this is falsifiable seems difficult for ID critics to deny, for then they would have to give up the claim that they are justified in saying that ID is not the best explanation for certain features of living things, and they would have to give up any claim of us knowing that some rival theory explains it better. Another formulation one could make is “the theory that intelligent causes are necessary to explain certain features of living things.” This claim also appears falsifiable (at least in the “it is possible in principle to have sufficiently strong evidence against it” sense); if we show that natural processes can do the job, one will have falsified this sort of ID theory.