My response to people in the newsfeed posting nonsense about ‘Sweden’ ‘it’s just like the flu’ or some other toxically ignorant para-numerical bullshit:

There’s a dangerously oversimplified (and contagiously common) ideological bias at play in the newsfeed among people who think ‘they’re presenting an alternative view’. Here’s the problem with their failure to understand damn near everything, let alone the situation they think they are commenting on: apparently, they suppose that ‘the same thing’ is happening ‘everywhere’, and that we can (and should) generalize from one situation to another.

Well, if that were true, everyone who was infected with the novel coronavirus would either be symptom-free, or have exactly the same symptoms, or die. But it would always be the same. Every time.

Notice how that isn’t what happens? Ever?

Good. You’re starting to understand that every local person and situation has unique features. It’s a start.

Unfounded generalizations derive from the misapprehension that a: there’s just one thing going on (there are thousands) and b: it always goes on the same way everywhere (this is formally impossible). While a modest degree of useful generalization is plausible (and that’s what we are basing recommended behavioral responses on), pretending that because you can cherry-pick some situation where lax response hasn’t yet led to apocalypse… and thus propose that everyone should return to life as (ab)normal is criminally negligent.

Now, let me talk about the unknown as it relates to the data thus far acquired and shared. There’s a vast universe of data out there (infinite data). When summed, the entire cache of data humans have acquired on this phenomenon is minute. We have a very modest (but useful) grasp on an extremely complex plethora of phenomenon. Most of this phenomenon is mysterious, unknown, a galaxy of question marks. It’s like 85% unknown. Humans don’t even understand why there is stuff at all, let alone what stuff actually is. We parade around as if human knowledge is complete (or nearly so, or approaching completeness), when it barely approaches reasonably partial.

This doesn’t mean that what we can learn isn’t useful; it is, if we use it intelligently. It means that the appearance or pretense of completeness… from anyone advertising ‘I know what’s going on’ is and must be totally wrong. No one knows what’s going on. What we have is enough partial information to make projections, and to respond to best evidence with preventative intelligence. And believe me, we had better do that, because if we do something else, the results will definitely be abysmal and terrifying.

Second, each unique place, situation, population, array of social and ideological behaviors… is unique. So you get significant differences in outcome from place to place (just as you do person-to-person, time-to-time, and so on). Keep this in your mind; generalizations emerge from summations of phenomenon over time, populations and situations. Any specific situation can (and will, eventually) vary dramatically from ‘the statistical norms’ that emerge from datatsets.

Thirdly, data science is incredibly complex. And data is collected, reported, evaluated, summarized and distributed differently in different places, times, situations… and ideological contexts. Red states may report differently than blue states. Some places have decent reporting, others have terrible reporting. We have partial testing, with some tests that have variable efficacy.

And statistics only give a tiny fraction of the actual situation, because you can’t get all the data, ever. And you can’t get it without biases, either. Worse still, data can never catch up to the actual situations (and these are always myriad) that it pertains to.

As if those stumbling blocks were insufficient, in this case, »our behavioral responses to the situation feed back »into the data. Now what you have is a situation where you can’t tell what the baseline is (or would be), because it was totally obscured by prompt, relatively intelligent preventative action.

And as that process »continually feeds back into the already relatively mixed-up and variable data, you get significant divergence all over the place… for precisely the reasons I have outlined above, not to mention the 10 reasons we haven’t yet discovered… and probably never will.