Traversing the Razor: Parsimony in Science

 

“Plurality must never be posited without necessity.”

-William of Ockham

 

People prefer simple solutions to problems. This is pretty obvious – it takes less unnecessary hard work. So, does this idea apply in science? Welcome down the rabbit hole of the history and philosophy of science.

 

Ockham’s Razor is a hugely influential heuristic (rule of thumb) in science. The Razor provides a way to decide between competing explanations that are equally supported by the evidence at hand. It suggests that the favoured explanation is that which posits fewer variables.

 

However, science is not often ‘simple’. How do we translate this position to science? This is where ‘falsifiability’ comes in, a concept made famous by Karl Popper in The Logic of Scientific Discovery. For a hypothesis or statement to be scientific, there must, in theory, be a way to disprove it. The famous example is “all swans are white”, which can be disproved by the discovery of a black swan. By inductive logic, no amount of white swans can prove this statement; it is supported until a single black swan is found.

 

Falsifiability alone does not, however, reduce possible explanations to one. Competing theories may all be falsifiable, thus scientific. Having established that swans can be black or white, we propose three competing ideas: 1. All swans are either black or white. 2. All swans are either black or white, but location determines which. 3. All swans are either black or white, but location and the season it is determine which. Having taken samples from Australia and England to test these hypotheses, you would see that all statements are supported, but number 2 has an extra variable, and 3 has two.

 

Strict application of the Razor would suggest you accept hypothesis 1; it is the simplest description of the situation. However, the extremely strong correlation between location and swan colour suggests that 2 is also acceptable. In this case, you decide that the ‘simplest’ hypothesis is the weaker, because it has less explanatory power. That is, even though clearly swans are either black or white (hypothesis 1), the black swans are all in Australia, and so hypothesis 2 suggests an explanation determined by geography. What about hypothesis 3?  Well, seasons are dependent on location, and so the seasons variable is superfluous, regardless of how well supported it is by the results. We take hypothesis 2 and move on, because that result has thrown up new hypotheses (e.g. around species and location) – the very fodder of science.

 

Thus science aims to explain, rather than simplify. Ockham’s Razor is really about how to prefer an explanation, rather than about the most simplistic explanation. Sometimes the best explanation is very complicated; the point is that it is no more complicated than it needs to be to do the necessary explaining. Many ‘conspiracy theories’ fall foul of the Razor for this reason; they introduce extra variables without improving the explanatory power. The non-conspiracy hypothesis can explain all the evidence.

 

Things get interesting when a contradictory result is found by a new experiment. Does it really falsify the hypothesis, or should we modify the hypothesis? Should hypotheses be ‘backward modified’ like this to explain new data? Doesn’t this contradict everything I’ve just said? How the hell does science really work? This will be a tale for another day, when we meet people like Thomas Kuhn and Imre Lakatos, and encounter the anarchist, Paul Feyeraband.

 

Michael blogs at traversingtherazor.wordpress.com