Thread by @doctorow

A common pathology among experts in quantitative field is to assume that when they are dealing with problems with qualitative dimensions that they can simply model the quantitative part and ignore the squishy human factors. This does not end well.

1/

This has cropped up a LOT in epidemiology during the plague months, for example in the belief that untested "exposure notification" with mobile apps can substitute (or even augment) shoe-leather "contact tracing":

pluralistic.net/2020/08/20/dub…

2/

But the Dunning-Krugerish insistence that quantitative methods can be applied to domains with strong qualitative components isn't limited to contact tracing - it also applies to re-opening plans.

3/

The University of Illinois decided to re-open and bring 40,000 students back to campus on the basis of models created by physicists who were dismissive of the entire enterprise of epidemiology.

https://twitter.com/yellingatwind/status/1301700119419432960?ref_src=twsrc%5Etfw

4/

"We learned to like epidemiology but I cannot imagine spending another 5 years doing it" - because it didn't provide the same intellectual thrill as physics, but everyone makes sacrifices, and for these physicists, they had to give up "intellectual curiosity."

5/

Their model predicted that the case-load would stay below 100, in a worst-case scenario. Instead the cases rapidly spread to 780 before the entire system went into lockdown. The cause? "Partying among students."

6/

That is to say, the model did not account for the qualitative, anthropological elements of the system it was modelling. The reason these qualitative elements were dismissed is that they're hard to model.

7/

It's basically a lethal version of the old joke about looking for your keys underneath the lamppost instead of where you dropped them, because the light is better there.

8/

Yes, yes, "All models are wrong but some are useful," but the predicted utility of your model is in inverse proportion to how many qualitative factors you have to elide, not because of their insignificance but because of their intractability.

9/

When you vaporize the qualitative elements, the quantitative residue that remains is of unknown usefulness.

Image:
Xinh Studio (modified)
thenounproject.com/term/physics/2…

CC BY
creativecommons.org/licenses/by/4.…

eof/