@uh

Two Questions: is this a great example?

1) Distinguishing direct, indirect and total effects is exactly what causal modelling in epidemiology seems to excel at, e.g., http://ftp.cs.ucla.edu/pub/stat_ser/r251.pdf

and dynamic causal modelling has been applied to complex systems to get at the point about dependence and independence at diff. scales:

http://philsci-archive.pitt.edu/16882/1/PSA-ND-Revisions-Unformatted.pdf

So what is being added here? 1/2

@uh

2) Is any of this really about mereological relations? Herd immunity seems a bad example to me in that, here, the population seems to be nothing over and above the set of all individuals and all relevant dynamics are dynamics between those individuals

so are the highlighted indirect and total effects really “top-down mereological effects”?

(note there is nothing “top down” in a typical causal DAG representing relationships between such variables, see first paper under 1.)

2/2

@uh @dcm@social.sunet.se @UlrikeHahn@fediscience.org what did you think of this recasting of the epidemiology stuff in terms of contextual and 'mereological' factors? It struck me as just a terminological variant that does not add much. I'm also a bit confused as she seems to be going into context-dependent constraints before she actually explains what they are, which is supposed to happen in the next chapter...

#JuarreroBook

@dcm@bookwyrm.social @UlrikeHahn @uh Just saw that you had already formulated similar questions over here that for some reason I missed on bookwyrm!

I fully agree about the case for top-down causation being rather weak with the epidemiology example, as those population-level dynamics indeed seem to be just a matter of inter-individual interactions (so nothing over and above them).

@dcm@bookwyrm.social @UlrikeHahn @uh in a sense, achieving herd immunity may change people's behaviour, but that's also the sort of normal causal relations we would expect: if people believe they are safer because of herd immunity, they will behave differently, etc., but then the cause is the belief, transmitted socially, etc., rather than the population-level property per se.

@dcm@social.sunet.se @dcm@bookwyrm.social @uh

A plausible example of the kind of emergent whole that's a candidate for downward causation is a traffic jam (TJ)-

TJs have properties like location, duration, and movement. These are arguably props. of the TJ not the indiv. cars (because e.g., the TJ's 'movement' is in the opposite direction of the individual cars & different speed (when viewed in time lapse..).

But pop. herd immunity strikes me as a short hand for a complex situation not a 'property'

1/2

@dcm@social.sunet.se @dcm@bookwyrm.social @uh
2/2
I'd argue we see this from the fact that a percentage (e.g., 95% = herd immun.threshold) is a simplification/approximation that ignores network structure. If all unvaxxed 5% are in a local cluster, they won't be protected just because the threshold has been met.

The true causal process unfolds across the contact network. So herd immunity isn't a causally efficacious property of "population" - it's a conceptual shorthand.