What Makes an Explanation Bad?

When it's easy to vary

In #articles

In David Deutsch's TED Talk he describes what makes good and bad explanations. In particular he outlines what makes scientific explanations good ones whereas mythological ones bad. Many people claim that what makes an explanation bad is its lack of testability - goblins are a bad explanation for my car breaking down because you cannot test for goblins. Although somewhat true, this is not the primary problem with the explanation, according to Deutsch. The problem is that, even if shown to be false, the goblin explanation is too easy to vary. We can tweak it slightly to get the different prediction. The example in the talk refers to the Greek gods creating the seasons, and that one can easily modify the presumed behavior of those gods to get any result you'd like.

In science, we look for explanations that one can't wiggle out of. If it is wrong, it must be discarded. The Higgs boson held a specific and irreplaceable role in the overall theory, and if not observed the theory would have to be seriously rethought and restructured. Some of the issues with string theory is that it predicts too much, and is too flexible, and thus fails as an explanation.

A very similar idea I had written about before, with respect to the accounts of the resurrection of Jesus. An explanation that is too flexible becomes meaningless, and shouldn't be relied on. It's the same idea that underscores overfitting in regression and neural network models and the role of Ockam's razor.