3 Simple Rules for Believing in Science
I saw Nacho Libre this past weekend. When Jack Black’s character asked his friend Esquelito if he believed in God, Esquelito responded, “No! I only believe in SCIENCE!” The audience burst into laughter.
What does it mean to believe in science? Some feel that Occam’s razor is the cornerstone of scientific thought. The way I make this concrete is through the following three rules.
- If one theory is simpler to describe and allows us to predict more than another, we favor it.
- Given two models that allow us to make the same inferences/predictions, we favor the model that is simpler to describe.
- Given two models that are equally easy to describe, we favor the model that allows us to make more inferences/predictions.
One example of the rules in action can be found by comparing the works of Tycho Brahe and Johannes Kepler. Tycho produced a model that, while incredibly difficult to describe, predicted the positions of celestial objects with uncanny accuracy. When Kepler introduced elliptical orbits and changed the frame of reference from the Earth to the sun, he managed to produce a model as accurate as Tycho’s but greatly simplified in its description. Thus, by rule #2, science accepted Kepler’s laws over Tycho’s.
In a recent conversation, a friend made the point that while scientific models can allow us to make inferences about the present and perhaps the future, this does not necessarily mean that the rules currently in play are ones that have always been. The point is significant since it is possible there may have been transient behavior in the past that has since settled down.
Our conversation was in the context of Darwinian evolution, which I will simplify into the following two parts: common ancestry and natural selection. As I mentioned in an earlier post, natural selection has allowed us to make models for predator-prey cycles and predictions about antibiotic resistance. My friend’s point should make one ask whether common ancestry provides any additional predictive or inferential power. If it does not, then rule #2 should cause us to favor a model that only has natural selection.
The main place where common ancestry has helped scientists make inferences is in phylogenomics. The mathematical models that form the basis of comparing DNA and protein sequences across species rely on the assumption that these organisms have a common ancestry. For instance, the application of tree Markov models is based entirely on this assumption. Further, these models have allowed us to infer genes and their functions in humans by comparing them to organisms whose genomes are better understood. Thus, it would appear rule #2 does not come into play in this case.
Of course, there is a possibility that someone may come up with a theory that does not have a common ancestry assumption but endows the same inferential/predictive powers as the current ones. If it also provides additional insights, either rule #1 or #3 could apply and exclude the models with common ancestry. Another approach is not to believe in science.