"You Will Never Get Causal Information Out Without Beginning By Putting Causal Hypotheses In."

[Literally]

"Pearl began his work on artificial intelligence in the 1970s with this mindset, imparted to him by his education. For much of the scientific community throughout the twentieth century, the very idea of causation was considered suspect unless and until it could be translated into the language of pure statistics. The outstanding question was how the translation could be carried out. But step by painful step Pearl discovered that this standard approach was unworkable. Causation really cannot be reduced to correlation, even in large data sets, Pearl came to see."

I thought this book review was interesting. Pearl would say just running lots of experiments won't work; one needs a model to test. I have been running into these "graphical models" in the literature more often... although I am still not there yet. What is interesting is how few people are. I recently spoke with a professor of machine learning who said that that graphical models seem like answer to formalizing causation, but he hasn't learned the math either. This article is the first thing I've come across that makes the idea accessible.

UPDATE: Guess it would help if I included the link

The Why of the World

 

[LeeLemonoil]

Your knowledge and stillset is very valuable here on RPF @Literally, pls keep up posting.

Remotely related to your OP but maybe I’m the same theme:
Have you any knowledge about the various models and simulations that climatology especially the anthropogenic-climate-change are based on?
How would Pearl regard those? Aren’t they necessarily only correlations presses into pseudo-causative models?
I‘ve read a newspaper article about Stevens Meta Review based on 20 climate-change models and possible outcomes differed greatly and such... it’s a polluted debate and asking questions quickly places one into climatechange-denier category.

 

[ecstatichamster]

The idea is to form an explanatory theory and test that through experimentation, that I get.

I don’t get the graphical representations stuff, but I do agree with the explanatory theory part. That’s why I am such a believer in Dr. Peat’s findings. He finds explanatory theories for things.

Experiments today are done for data gathering and without regard to explanations, and this is very deeply flawed. The theory of induction itself is just a tool. What makes the tool work is a thinker who has an explanatory theory.

David Deutsch argues this very well in his books. It is all but forgotten today when empirical data gathering seems to be an end in itself, and explanations for things are largely ignored or culled from the data. Either interpretation approach is deeply flawed but it suits Big Medicine. They can more easily manipulate statistics than they can explanatory theories, and they just come up with a stupid explanation (SSRIs work by Increasing the “good” serotonin, women need estrogen replacement etc.) and sell drugs.

 

[Lejeboca]

Thank you for posting this link, @Literally.

That’s why I am such a believer in Dr. Peat’s findings. He finds explanatory theories for things.

That what I like too in Ray Peat's approach. In addition, haidut's assertion that it is important to see how an experiment was conducted not just what data (results) it collected.

Experiments today are done for data gathering and without regard to explanations

In William Blake's "The true method of knowledge is experiment" , the gist, I think, is that the experiment is a method for knowledge and not the end-in-itself or not what happens with (massive) data collection --- just in case we will need this data for something.

 

[Literally]

@LeeLemonoil thank you.

Have you any knowledge about the various models and simulations that climatology especially the anthropogenic-climate-change are based on?
How would Pearl regard those? Aren’t they necessarily only correlations presses into pseudo-causative models?
I‘ve read a newspaper article about Stevens Meta Review based on 20 climate-change models and possible outcomes differed greatly and such... it’s a polluted debate and asking questions quickly places one into climatechange-denier category.

My direct knowledge about climate models is really nothing. I know about some of the related maths / simulation tactics. That would be an interesting question for Pearl... I plan to read his book at some point. In searching for climate change publications that mention graphical models, they are a few but I didn't see any of the kind you are referring to.

Yes I think they are probably butchering the correlations in many cases. From talking with a friend who studies such things, I think there are problems at multiple levels. For example, the heat island effect which raises temperatures around cityscapes. If you look at the proportion of data sources that could reasonably affected by heat islands, it has grown along with urbanization. At a deeper level many of the simulations they use involve chaotic effects which make predictions very difficult.

There is a woman recently who came out with a dynamo model of the sun that seems to provide an elegant formula (with double cosines) that explains much of our temperature records... have you seen that? I can post it here if it interests you.

 

[Literally]

@ecstatichamster

I don’t get the graphical representations stuff, but I do agree with the explanatory theory part. That’s why I am such a believer in Dr. Peat’s findings. He finds explanatory theories for things.

I think the first step to understanding "graphical" models is to realize they mean this kind of graph -- Graph (abstract data type) - Wikipedia

These "graphs" do not have to be represented visually, the point is they can capture all the possible connections. As with a lot of scientific theories, the really interesting part is probably int the details of the math rather than this basic notion of, "hey, let's formalize expressing all the possible ways causation could be wired up, and get people talking about whether or not they have properly explored that space," which is a formalization of something scientists have always done.

Experiments today are done for data gathering and without regard to explanations, and this is very deeply flawed. The theory of induction itself is just a tool. What makes the tool work is a thinker who has an explanatory theory.

Well said.

David Deutsch argues this very well in his books. It is all but forgotten today when empirical data gathering seems to be an end in itself, and explanations for things are largely ignored or culled from the data. Either interpretation approach is deeply flawed but it suits Big Medicine. They can more easily manipulate statistics than they can explanatory theories, and they just come up with a stupid explanation (SSRIs work by Increasing the “good” serotonin, women need estrogen replacement etc.) and sell drugs.

I think Deutsch is beyond brilliant... but I haven't penetrated much of his physics work. Impressive that you are reading him.

@Lejeboca

In William Blake's "The true method of knowledge is experiment" , the gist, I think, is that the experiment is a method for knowledge and not the end-in-itself or not what happens with (massive) data collection --- just in case we will need this data for something.

Also very well said, thank you.