I’m reading the book Lost in Math by the famous Youtube content creator and scientist Sabine Hossenfelder.

The author shares stories around a common theme – beauty shouldn’t be an argument in science. According to Sabine Hossenfelder, scientists have the tendency to chase beautiful ideas and dismiss hacky, unnatural solutions that explain the world very well due to their ugliness. She gives the heliocentric model as an example. Scientists from the past had difficulties accepting that stars are as far as we know they are because the numbers were too large, which felt unnatural. I’m encountering this type of problems relatively often, with beauty being used as an argument for expensive ideas, for example new standards, or second systems / rewrites.

Who could’ve imagined that this is where I’ll find cues on my quest on figuring out web experimentation (yeah, I’ve not blogged about that, but maybe it’s a good moment to start). I’ll just share two quotes.

…you could go most of your life without having to confront [the results of an] experiment”

Apparently, physicists sometimes run experiments that take a really long time, like decades. For example, particle accelerators that are expecting to produce a new particle in decades, meanwhile postponing decision making and facing the realities of known models not producing the expected results. In my world, where experiments are run on web, I’ve also encountered long-running web experiments. Usually because the sample size is small or because we expect some distant future event to happen, which never comes.

“If you are an honest physicist, 99.99 percent of your ideas, even good ideas, are going to be ruled out, not by new experiments but already by inconsistency with old experiments.”

Same applies to web experimentation as well, although the share is probably not 99.99% (as it isn’t for theoretical physics either). Verifying ideas against prior experiments can filter out many of them quickly. But there’s always the problem – does that old experiment still apply even if we ran it X years ago? What if something in the setup was wrong and it works now?

This book touches my chords despite being unrelated to my work. Something in the area of process is overlapping. I don’t know why yet but I’ll keep reading.