Jim Al-Khalili has an article at OpenMind attacking Occam’s razor, at least in the form it’s typically articulated, that the simplest explanation should be preferred. Al-Khalili correctly points out that there are a lot of problems with that version of the principle. Simply preferring the explanation we think is the simplest is often just favoring the one that flatters our biases. Taken to extremes, as it often is in political discourse, it leads to impatience with complex subjects and the expertise necessary to make judgments about them.
But I don’t think the version of Occam’s razor he’s attacking is the one actually proposed by William of Ockham. We can see this if we go with a more precise description of it:
entities should not be multiplied beyond necessity
“Entities” here represent assumptions or components of an explanation. But the key portion is the final two words: “beyond necessity”. It’s not the simplest explanation we’re looking for, but the one with the fewest assumptions that still fits the data, that correctly predicts observations.
Al-Khalili sort of recognizes this when he notes that his preferred interpretation of Occam’s razor is that we should prefer the more useful theory. I’m on board with that. However we might have multiple theories that are equally useful, but typically some will have fewer assumptions than the others. Each additional assumption is an opportunity to be wrong, so the theory with more of them is more likely to be wrong than the theory with fewer.
It’s worth noting that there is never just one explanation for observed phenomena. There are always alternatives. Theories like general relativity, quantum mechanics, or natural selection are often the simplest among those alternatives. We can always posit more complex theories that make the same predictions, such as theistic evolution, the equivalent of saying that thunder obeys the laws of electricity but Thor or Zeus is still around making sure of it. Scientists generally don’t favor these alternatives because of the additional unnecessary assumptions.
Of course, as Al-Khalili notes, Occam isn’t a law. It isn’t guaranteed to be right. It’s just a useful heuristic. But one that I think history shows is very useful.
One of the examples Al-Khalili discusses is Copernicus’ famous argument that Earth orbits the sun rather than the other way around. Often Copernicus’ theory is presented as though it was vastly simpler than the reigning Earth-centered Ptolemaic theory. But Copernicus’ theory, for someone in 1543, was not obviously simpler than Ptolemy’s. Astronomers did quickly recognize its mathematical elegance, so it was simpler in that fashion, but its ontology seemed just as complicated. And for most people at the time, it seemed utterly bizarre.
Copernicus kept in his model the ancient conception of the planets moving in perfect circles, as well as being embedded in concentric crystalline spheres. It was these assumptions that forced him to retain his own version of epicycles. In other words, while he was less wrong than Ptolemy, his cosmology still had a lot of issues. But from his perspective, these were necessary elements of the model, needed to explain how the planets moved in the correct manner.
Al-Khalili’s point is that the eventual correct explanation, involving ellipses and gravity, was far more complicated than what Copernicus envisioned. And that’s true, but that doesn’t mean Copernicus was in any position to assume those more complicated explanations. If he had tried, it would have been an exercise in wild speculative guessing. Those guesses could only have hit on the right answer by sheer chance, an extremely slim chance given all the alternatives.
What Copernicus needed were Tycho Brahe and Galileo’s observations, which wouldn’t be made for several decades. Tycho’s observations would show comets crossing where the crystalline spheres were supposed to be, indicating they didn’t exist. Johannes Kepler, with Tycho’s data in hand, was eventually able to replace the circles with ellipses. And Isaac Newton, with all the telescopic data available by the late 1600s, was able to explain the motion of the planets with a theory of universal gravitation.
But Copernicus was in no position in the early 1500s to do anything along those lines, no matter how much of a genius he might have been. He just didn’t have the data, the evidence that would force the necessary paradigm shifts. All he could do was use parsimony, Occam’s razor, to put forth the least wrong theory possible with the data that was available.
It’s a sobering reminder that we may not even realize some of the assumptions we’re making. The assumptions might be implicit in the current paradigm, one we may not even realize we’re in. But it doesn’t mean we should make even more assumptions.
Which gets us back to the key points. Occam’s razor is not simply preferring simpler explanations, but favoring explanations with fewer assumptions that still make accurate predictions, that fit the current data. In that sense, it’s a useful heuristic, not a law guaranteed to produce the right answer, but one that helps us avoid answers more likely to be wrong. And that makes it a useful tool, one recognized by people like Isaac Newton and Albert Einstein for its value.
What do you think? Am I missing some problems with Occam’s razor? Or any benefits from tossing it?