Ishi Crew
Ishi Crew
Sep 2, 2018 · 3 min read

I came across this set of papers because a FB group on ‘philosophy and foundations of logic. math and science’ (or something like that) linked to it yesterday (its an international group and i think the people are a mixture of students, amateurs — some who have their own theories who in sciences might be called the ‘lunatic fringe’, and a few scientists/logicians — -many of whom work in very specific fields so the FB group for them is really just a place to learn about things outside their own field — somewhat like Science magazine).

I was surprised to see that Judea Pearl was mentioned — -i had just downloaded from his web site the 1st chapter of his new book ‘the book of why — science of cause and effect’. (actually i was only surprised because his name was mentioned, because many articles do not cite sources for various reasons — -economy, attention economy, hirsch index, etc. ).

I actually was looking at that because similar ideas were being discussed on John Baez’s Azimuth blog — petri nets — -and the place i first heard of those was in a book by Pearl — so I mentioned Pearl on Baez’s blog (which is a very good one). (I also learned that Petri nets were originally devised to understand chemical reaction kinetics — i have a degree in chemistry though i was actually into theoretical biology but they didnt have degrees in that so i called it chemistry — -mostly took physics and a little biology and the required courses for chemistry).

Also Heckman came up — -i’ve read a few of his papers — which were on same topic though my view is econometricians have their own dialect which i am not very fluent in.

And spurious correlations also came up — -the yule-slutsky effect or theorem on spurious correlations and ‘moving averages’ from 1930’s is one of my favorite theorems. (I first learned about that in some old book on ‘math methods in social sciences’ and some old papers in Am Math Monthly. There is a great linguistics/evolution blog that has alot on spurious correlations called replicated typo.)

The equations in the above paper are very similar to ones used in statistical physics/stochastic processes , which overlap with ones used in quantum mechanics. The way I learned statistical physics, we never heard of Bayes theorem — — we just had ‘master’, diffusion and fokker-planck equations, and markov and or non-markov processes (following N G van Kampen, a famous physicist, i call non-markov processes ‘higher order markov’ processes (ones with long memory). ( I learned what little I know about these subjects from old 1960’s books and articles by van Kampen and Mazur — -not barry Mazur — another great mathematician — -but P Mazur from europe.)

‘Selection bias’ is a very interesting issue. I think it comes up in issues like whether certain groups are under or over represented in academic fields — eg females, african americans, asians, GLBTQ+ types, income and geographical groups, family pedigrees, etc.)

I think it also comes up in quantum theory — -the so-called ‘detection efficiency ‘ or ‘fair sampling’ problem (see https://en.wikipedia.org/wiki/Loopholes_in_Bell_test_experiments ) . Its like example given in your paper — -if you conduct a political poll by just calling people on a smartphone, you may not get a ‘representative sample’.

I do wonder if the seeming fact that I came across this paper which mentions Pearl and Heckman the same day i had been reading this sort of stuff stems from a ‘common cause’ or is just a spurious correlation. I am sort of into ‘counterfactual reasoning — -eg Banach-Tarski paradox — -or how to turn a sphere inside out without cutting it. So I can conceive of situations in which a correlation is a cause.
(The book by physicist R F Streater ‘lost causes in and beyond physics’ is where i learned the term ‘common cause theory’. Its similar conceptually to your diagrams on ‘disaster, alarm, traffic’ . I think noble laureate physicist g t’hooft has another ‘common cause theory’ for quantum mechanics , although he may have 2 interpretations — one deterministic, the other probabilistic. I see 2 such interpetations)

I am working on my own version of causal nets (which i likely wont finish) but its very abstract , relies on number theory theorems i cant prove — hardy and ramunajan proved those, and is supposed to be very simple — -and has its own diagrams similar to ones Pearl uses. Its like a recipe.

I mostly deal with social issues like resource allocation including time and money — or utility maximization — but its all nonlinear.