We’re pretty bad at theories, it seems, because we don’t really look for disconfirmation.
In Dan Gardner’s book Risk, he recounts an experiment done to show this that was conducted by Peter Watson.
The challenge is pretty simple: given 3 numbers in sequence, can you figure out what the rule is? Participants were allowed to write down 3 different numbers to see whether they followed the rule, and try this as many times as they wanted.
So here are the numbers: 2, 4 and 6.
It seems pretty normal, so most would then ask the researchers whether these numbers fitted the rule:
8, 10, and 12. And yes, they do.
And if they wanted more vigorous testing, they would ask whether the following sets of numbers followed the rule:
14, 16, 18 or 100, 102, 104. And yes, both do follow the rule
So what’s the rule? Well most said that it was “any 3 even numbers ascending by 2 each time”. And they were wrong. That’s not the rule. The correct rule is: “any 3 numbers in ascending order”.
What had happened, it seems, is that people didn’t try to disconfirm the rule. They didn’t ask, for example, whether “3,4,5” followed the rule.
As Dan Gardner says,
“most people do not try to disconfirm. They do the opposite, trying to confirm the rule by looking for examples that fit it. That’s a futile strategy. No matter how many examples are piled up, they can never prove that the belief is correct. Confirmation doesn’t work”
Something I need to bear in mind while trawling the myriad posts on Everything 2.0. It seems it’s better to look for indications that I’m wrong rather than bask in the warm webby-goodness of confirmed 2.0 successes. And intuitively that makes sense. Rigour is surely preferable to comfort.
Yup. This reminds me of Karl Popper’s concept of falsification, which he put forward in his Logik der Forschung (The Logic of Scientific Discovery) in the early 1930s. Rigorous inquiry should be based on a constant process of attempting to disprove (falsify) theories, rather than attempt to validate them through confirming instances.