On a number of occasions, I’ve encountered climate skeptics invoking Occam’s razor to claim that because their idea/model/theory/analysis is simpler than someone else’s that their’s is therefore correct, and the other is consequently wrong. Firstly, according to Wikipedia,
In science, Occam’s razor is used as a heuristic (general guiding rule or an observation) to guide scientists in the development of theoretical models rather than as an arbiter between published models. In the scientific method, Occam’s razor is not considered an irrefutable principle of logic or a scientific result,
which indicates – as would be obvious to many – that Occam’s razor is not a suitable way to differentiate between different models.
Occam’s razor, however, is a perfectly sensible idea. If you want to understand something, it is best to make your model/theory/analysis as simple as possible. You shouldn’t include things that aren’t necessary or that aren’t physically motivated. It doesn’t, however, mean that you can – for example – ignore an important bit of physics simply because you think it would make your model too complicated. Also, it’s not something that I’ve ever seen a credibly scientist use to justify why their model is the best possible model.
There are two examples where I’ve seen it used by climate skeptics. One was an occasion I’ve discussed before (Watt about 6 degrees?) in which someone assumed that CO2 concentrations were increasing linearly with time (which a quick look at the data will tell you is not a particularly good assumption) and then concluded by saying that because this analysis was the most simple it therefore was the most likely. The other was someone who said that because the trend in the temperature anomaly data since the mid-90s was not statistically significant (i.e., it could be positive or negative) that Occam’s razor tells us that the most likely trend is 0oC per decade. No it doesn’t. The actual data analysis tells us what the most likely trend is, not some guess based on a philosophical construct that is simply intended to guide us when developing models or carrying out some data analysis.
There is a somewhat tongue-in-cheek law called Godwin’s Law, that I can’t explain because that would immediately invoke Godwin’s law and reduce the credibility of everything I’ve said in this post. It’s existence, however, has made me wonder if we shouldn’t add a corollary to Occam’s razor which says that if you need to invoke Occam’s razor to make your model seem the most credible, your model immediately becomes the most complicated of all possible models and, therefore, by Occam’s razor is then the least credible of all possible models.