Call for resources on the link between causation and ontology

I’m a lit­tle ashamed to ad­mit I only read “Why Cor­re­la­tion Usu­ally ≠ Cau­sa­tion” yes­ter­day. It’s very, very good, and you should read it too.

My es­sen­tial take­away from it is this: You can find nonzero cor­re­la­tions be­tween al­most any­thing you care to mea­sure. How­ever, it seems un­likely that the num­ber of causal re­la­tion­ships in the uni­verse scales at all pro­por­tion­ally to the num­ber of cor­rel­a­tive ones in the Uni­verse.

This ques­tion feels like the wrong one to be ask­ing to me, some­how. It feels on­tol­ogy-fla­vored, in a way that doesn’t make it a great match for how I nor­mally think about statis­tics, and I would ap­pre­ci­ate some book recom­men­da­tions on the sub­ject in the com­ments. But first, let me try to ex­plain my think­ing on this.

Start with the “base” layer of re­al­ity, the move­ment of atoms, or elec­trons, or strings, or what-have-you. If we are watch­ing the ac­tions and re­ac­tions of that layer from afar, then it seems to me that we have the best pos­si­ble en­vi­ron­ment for do­ing a few ex­per­i­ments to first demon­strate cor­re­la­tion, and then a few more to demon­strate cau­sa­tion af­ter­wards. While we can never be 100% sure, we can asymp­tot­i­cally reach cer­tainty in that world. So far, so good; there’s a rea­son ex­per­i­men­tal physics can get so pre­cise with its pre­dic­tions.

When you go one layer of ab­strac­tion up—to molecules, if our base layer was “atoms”—it seems to me that sud­denly the difficulty of as­cer­tain­ing cau­sa­tion should sky­rocket. There are many more con­found­ing vari­ables and pos­si­bil­ities, that make de­sign­ing an ad­e­quate ex­per­i­ment much harder. In ad­di­tion, it is harder to define “molecule” pre­cisely than it was to define “atom”. How far do we move the con­stituent atoms apart be­fore we turn a molecule into a non-molecule, for ex­am­ple? That seems like a ques­tion that you have to some­times an­swer differ­ently de­pend­ing on differ­ent sce­nar­ios.

The ex­per­i­ments you run for cor­re­la­tion be­tween molecules, on the other hand, might be harder, but I don’t get the feel­ing they ex­pe­rience the same kind of… Com­bi­na­to­rial-su­per­ex­plo­sion-y ad­di­tional difficulty that an ex­per­i­ment de­signed for cau­sa­tion has to han­dle.

You should prob­a­bly try to ac­count for things like ther­mal noise, and trace im­pu­ri­ties and the like, if you have to, but past a cer­tain point it’s sort of okay to let go of the reins and just say “We can do more cor­re­la­tion tests later”. The claim un­der­neath that claim be­ing that, those things which muck up the data are mostly due to ran­dom chance, and if we do the ex­per­i­ment again un­der differ­ent con­di­tions, we will get a differ­ent set of ran­dom cir­cum­stances wrap­ping around the ex­per­i­ment.

This prob­lem feels like it re­curs ev­ery time you go up a level, which is why it con­cerns me so much. When you get to the level of deal­ing with hu­man be­ings in medicine, it feels to me the difficulty of de­ter­min­ing cau­sa­tion must be so vast as to be al­most not worth the effort; and yet, at the same time, that in­tu­ition feels clearly wrong, be­cause there was a lot of low-hang­ing fruit in the world of medicine—vac­cines be­ing the ex­am­ple par ex­cel­lence. But on the other hand, vac­cines op­er­ate on a rel­a­tively sim­ple causal mechanism! Maybe it shouldn’t be sur­pris­ing that such low hang­ing fruit ex­ists; what would be truly im­pres­sive would be if we found an easy cure to some dis­ease founded upon prin­ci­ples that only show them­selves at the level of rea­son­ing about hu­mans them­selves, the same way we usu­ally rea­son about molecules-as-prim­i­tives in­stead of atoms-as-prim­i­tives when we start to do biochemistry

I apol­o­gize if this isn’t a ter­ribly clear ex­pla­na­tion of what I’m get­ting at. If any­thing in here strikes you as similar to a prob­lem you have thought about your­self and have read up on, let me know. At the least, I should be able to come back within a few month and be able to prop­erly pose my ques­tion.