Causality, Interaction, and Complexity | Hacker News Reader

Causality, Interaction, and Complexity

alexahn.com

50 pointsalexahn13 comments

I highly recommend the book Causality and Modern Science by Mario Bunge for a thorough and scientifically grounded treatment of the philosophical problems of causality and determinism. This article seems to be going in an interesting direction with respect to those ideas but I can't quite follow it.

Seconded.

As for the article, it's just a shower thought. I don't fault the author for choosing to sketch down little sparks of personal insight on a blog instead of in a private journal, but I don't know why someone submitted something so undeveloped to HN or how it kept some footing on the front page.

It makes me wonder if people might just enjoy seeing clusters of meaty philosophy and systems buzzwords placed near each other in valid sentences.

People don't generally read hacker news because of its polish, but because of its novelty. There's plenty of polished backwash to read elsewhere.

For me, undeveloped personal insight is just the kind of content I find most interesting here.

> People don't generally read hacker news

Twisting your words a bit, but it’s been my experience that “indeed”, people don’t read the articles here. Lots of upvoting from just the headline or a quick skim. “I don’t get it but I’m curious what the discussion will reveal for me if I upvote”.

The guidelines here don’t really have a problem with this attitude either. You can make an uninformed upvote, just not an uninformed comment. In practice, both are somewhat common.

His other posts are like that too. He could avoid writing these superficial posts if he read a few books first.

Weird idea. It first denies causality by claiming it's impossible to observe (what is obviously ridiculous to state as a general rule), and then tries to redefine it as an emerging property (where does it emerge from if you have no ab-initio rules?).

That idea probably appeared from a very specific context. It possibly makes sense on that context. But generally stated like that, it makes no sense.

The idea that causality is impossible to observe goes back to Hume. Hume thought that people notice two events happening one after the other and infer causation. But, according to Hume, there is no way to prove casuation and our doing so is just a habit of the mind.

That blank dismissal of everything just because we can't have any kind of conclusive proof is useless. It may have been important at the time where people were trying to prove God exist and etc, but it's not something you want to hold nowadays.

The fact is that all of the settled science is composed of causal models (non-causal models aren't useful), and we have some damn good confidence on some of those. Proof is not necessary.

Oh, and those causal models are usually ab-initio, so for whatever chaotic system appear out of them, they are the cause, not consequence. Again, a few aren't, and the article is clearly about some set of those few. But you can't just blanket throw away all causality.

Every now and then I see something like this upvoted on HN and wonder, "What the heck?" Feels odd to see something get to the front page with 23 votes and 1 comment when I can't understand what it's trying to convey. [Especially that ratio of votes to comments seems pretty rare, particularly on more confusing/philosophical posts.]

I'd love if anybody upvoting this would spare a thought on what you're taking away from this.

He is describing the difference that exist in dynamical systems (physics, computer systems) between microscopic and macroscopic descriptions.

If you have few interactions among things, you can analyze their interactions through the lenses of cause and effects. That's the microscopic part. Think about a snooker table, if it's normal sized and a normal number of balls, you can imagine more or less what will happen once you hit your ball.

Once you have many things that interact between each other, with each interaction going in different ways, tracing causal links becomes impractical. Things start to get "chaotic", in the sense that small variation can cause long term changes in their trajectories (this can happen in any system with 3 or more degrees of freedom and feedback, because 3rd degree differential equations can have two curves/solutions that coincide on one side and then bifurcate later). Given enough things, it is like playing snooker with an infinite table and infinite balls, and infinite players. Then it start to make sense to track the macroscopic values that are conserved in large parts of the infinite table, e.g. energy levels of the balls (kinetic) and the force and general direction with which each average player hits the balls (acceleration).

Does this help?

I didn't upvote, but maybe this helps.

We work in an industry that is people building systems-of-systems-of-systems, which seems to match the popular misconception of entropy.

People are desperate for viewpoints and tools to help manage that complexity.

As someone who has just enough science background to be annoyed by the misunderstanding of entropy, and enough architecture experience to be annoyed by the misunderstanding of how to effectively try to manage complexity, I get your frustration.

Error twords building simple systems that are easy to replace with defined boundaries and the org structure to build such systems is a better path.

Proving causality in complex systems is hard, if even possible, but if someone finds value in this model, the use of flawed analogies is probably useful for them.

Hopefully they realize this is a lens and they will need other lenses with models flawed in different ways to see the bigger picture.

As most complex problems only allow for approximation, I have tried to let go of what specific lenses people leverage and have moved to asking them to use multiple views.

At least in this case where invoking properties of macroscopic qualities, Gibbs free energy etc...

It is technically wrong in multiple ways, but I can see how people may find value in it