Thread by @tomaspueyo

Analogies have a bad rap vs. 1st principles. But both have a role to play.

When should you use Analogies vs. First Principles?

It depends on the novelty-complexity relationship.

The more you've seen a situation, the more you can predict what will happen next without understanding it. The ancients might not have known how the solar system worked, but they knew the sun was going to come up tomorrow.

So you can use analogies when you've seen something a lot and it's very replicable.

That's how a lot of the medical sciences work today, for example. We're not sure how things work, but if they work over and over again, let's keep doing them.

It's also why you can cook without knowing much about cooking, just following a recipe. It's predictable. It's repeatable. You can use analogies.

That breaks down in novel situations. For example, when there was a solar eclipse in ancient times, when a new virus appears, or when you want to create a new dish.

In these situations, the only way to predict what will work is to really master the fundamentals. To reason from 1st principles.

Neptune was discovered not through observation, but math.
earthsky.org/human-world/to…

When you want to create electric cars that work or rockets that are reusable, these are novel situations. You must reason from 1st principles to predict what will work. That's why Elon Musk is so much into 1st principles.
waitbutwhy.com/2015/11/the-co…

If a situation is very very complex and you don't understand it, you can't reason from 1st principles. But if some things have already worked in the past, you can use analogies, even if you don't get all the nuances.

Finally, if something is completely novel and you don't understand how it works, it's simply impossible to predict. Don't try. For these situations, create a resilient system, one that can deal with unforeseen setbacks. If it's antifragile, even better.
en.wikipedia.org/wiki/Antifragi…

In your discipline, figure out where you lay in the novelty-complexity spectrum and that will tell you what kind of research you need to do to predict outcomes.

Back to the top

https://twitter.com/tomaspueyo/status/1421076378678702082?s=20