Where is science used?

In 2023 idea that everything needs to be data led and backed in science seems to be reaching near-religious levels. This post isn’t intended to be a book burning exercise, rather a brief investigation into what science is and why it’s not always appropriate.

What is science?

I'm describing science as the application of the scientific method over time. The scientific method works on falsification - asserting things which are not true. We build a mental model of the world (known as a hypothesis) avoiding those things which we know to be false. We find ways to generate supporting or detracting evidence via experiments.

Philosophically this falls into a branch of epistemology known as “induction”. Induction is not without limits!

Problem #1: Where do hypotheses come from?

Mathematics has an advantage over science in that its knowledge is collected deductively from a timeless base. To give an example of this base, the word “between” describes a relation of three concepts - something in the middle of two other things. There’s no other possible configuration for this idea and at least three things must exist for the word to have any meaning. This is known as a universal and we see these universals as a-priori (practically speaking, eternal ideas). From these a-priori universals we can deduce many things like counting, trigonometry and topology.

The inductive reasoning of science is different; we watch to see if things behave as our model of them predicts, and if they do we increase our confidence in the model. An example of this is that every day of your life the sun has risen, therefore it will almost certainly rise tomorrow. We’ve collected a set of evidence across various instances of an event to establish a pattern of the sun rising at that particular time.

A hypothesis is that model built by us to describe those patterns, which helps us predict how those things will behave going forwards. This creates a bit of a chicken and egg problem however; how do we decide what our hypothesis is when with the same starting data an infinite number of hypotheses can be made? I could believe that the flying spaghetti monster is pushing the sun around the earth every 12 hours and my friend Bob could believe it’s the flying spaghetti monster pushing the earth around the sun every 12 hours and both are valid predictive models of reality. We like to think there is only one "valid hypothesis" at a time, e.g. dinosaurs were killed by an asteroid or the moon was a part of earth but there's nothing intrinsically separating two hypotheses that both match the current evidence. We only rank them on "elegance" and as they start to consistently make correct predictions. This ambiguity of science is often overlooked, though it exists to this day and always will exist.

"it seems probable that most of the grand underlying principles have been firmly established"

Albert Michelson, 1894 (two years before radioactivity was discovered).

None of this is intended to say science is useless and mathematics is somehow pure. Induction is essential to believing in reality at all. I hope to persuade you only that science and mathematics have quite different views of knowledge and that these differences are either more or less useful based on your intent.

Problem #2: Can a hypothesis really be proven?

Of course after a hypothesis has been tested in many different ways we come to describe it as a theory, indicating we are relatively sure it is true. For science to progress however everything must be open to scrutiny at any time - because it is validated inductively through falsification.

Every theory comes implicitly with this idea of sureness; that some theories are “more certain” than others – though we base this on a subjective analysis of evidence. In a very real way we choose to believe in a particular hypothesis in much the same way someone may choose a religion based on which has the most compelling literature.

Some of these theories are very reliable but still discovered to be wrong. Newton’s gravity was accepted as the way to describe gravity for around two hundred and forty years before being disproven. Likely 99.9% of scientists believed it to be beyond discussion during this time.

“The reasonable man adapts himself to the world: the unreasonable one persists in trying to adapt the world to himself. Therefore all progress depends on the unreasonable man.”

George Bernard Shaw

We should pause here though to establish that believing even in Newton's gravity is a useful tool to stop you falling to your death.

Problem #3: The replication crisis

When not busy with apples in the garden Newton was investigating light by jamming spoons in his eyes in the 1600s. In 2010 the CERN Large Hadron Collider (LHC) was switched on at a cost of $5bn, involving >10,000 scientists and over 100 countries. Putting spoons in your eyes is a considerably more straightforward proposition than building a massive particle accelerator; every time the spoon goes in something funny happens to what you see. The LHC took years to produce understandable results and is affected by everything from the ambient temperature, air pressure to background radiation sources. The difficulty is in isolating the thing you’re trying to build a hypothesis for. Things become increasingly difficult the smaller or faster things get (e.g particle physics) or in complex systems like the stock market.

"the future truths of physical science are to be looked for in the sixth place of decimals."

Albert Michelson (a partial save from the earlier facepalm)

While background radiation is an obvious problem for the LHC we see the same class of problem in other science-like decision systems. During the COVID-19 pandemic everyone seemed to be reporting revenue growth due to product decisions despite the obvious backdrop of people being required to stay at home, disrupting regular routines and replacing them with “screen time”. It’s very hard to distinguish macroeconomic trends from your average A/B test because we can't build an underground version of the product isolated from the world as is done with the LHC.

Product managers at amazon saying they're crushing it (surely product managers at amazon were claiming this as a neverending win of A/B tests before the bubble burst)

In conclusion

Science is great, it's the best tool humanity has to understand the world (by a large margin) but we risk damaging it by using it incorrectly. The incorrect use of science also produces garbage - a great product is not the sum of all successful a/b tests.

I think that it'd be better for the world if we always acted like science was a thinking tool in the same way using a stick to get honey out from a bee hive is a physical tool. I feel too that we should find more ways to acknowledge that at best we're sophisticated typing monkeys zooming through space on a ball of hot rock, rather than eventually omniscient ghosts in flesh bots hailing the scientific method.