Against falsificationism
In The Logic of Scientific Discovery Karl Popper proposed that falsifiability solves both the problem of distinguishing science from pseudoscience1 and the problem of induction, the observation dating to Hume if not before, that you cannot justify generalizations with the same logical certainty that you can justify the conclusion of a syllogism. Even though you’ve seen the sun rise every morning as have millions of people before you, it is possible (even if highly improbable) that it won’t rise tomorrow morning.
Consider a simple syllogism, this one an example of a pattern known asmodus ponens:
All human beings are mortal.
I am a human being.
Therefore, I am mortal.
The defining characteristic of a logically valid syllogism is that if the premises are true—“All human beings are mortal.”, “I am a human being.”—then the conclusion must be true, where “must” has the same logical force as the assertion that “equals” does in 2 + 2 = 4.
Your belief (and mine) that the sun will rise tomorrow morning is justified, but our belief doesn’t carry with it that degree of certainty, because we can imagine that the sun might not rise tomorrow morning. Earth might explode in the next 12 hours.2 So no matter how many times the sun has risen before, it might not rise the next time.
Falsifiability solves the problem of induction by arguing that inductive reasoning should not be the goal of science. What science can establish with certainty is which hypotheses are false. Popper argued that science advances by excluding more and more false hypotheses. And he argued that pseudoscience is characterized by theories unconnected to data or observation or connected only through special pleading, paying attention only to some of the evidence.
The logic of falsification is simple. Consider the hypothesis “All swans are white.” If that hypothesis is true, we can construct the following syllogism.
All swans are white.
This bird in front of me is a swan.
Therefore, the bird in front of me is white.
Well, if you’re a birder or you’ve been to souther Australia, you know that there’s a species of swan there, Cygnus atratus that is black. Thus, we can conclude with logical certainty that the hypothesis, “All swans are white,” is false. That’s how falsification works.
But there’s a problem with what I just wrote. Given the structure of the argument, if the bird in front of me is black, we know that at least one of the two premises must be false, but we don’t know which one. Even worse, it’s possible that I wrongly asserted that the bird in front of me is black. Maybe it’s a white bird that’s been dipped in black dye. There’s nothing in the logical structure of those premises that tells us which of those possibilities is right, only that one of them must be. We haven’t advanced nearly as far as we thought.
We depend on our background knowledge to judge whether our observation is wrong and whether one or both of our premises is wrong. More precisely, as a knowledge community, the group of people interested in and expert in the subject of our hypothesis use their shared background knowledge and experience to determine the appropriate conclusion, possibly after a lot of argument and additional observation.
Falsification isn’t as simple as people often think.