Total Pageviews

Monday, June 8, 2020

COVID-19 Fatality Models Shocked the World

Forecast models 
have been used for 
government decision
making for decades.

Forecast models created 
great public awareness 
when used to over-predict 
COVID-19 deaths.

COVID-19 exposed 
the problem with 
computer models.

Computer models of
a complex system 
rarely explain
something in 
the real world
well enough to 
make predictions.

But they get abused 
by policymakers
anyway.

Computer models
are actually a gross 
simplification of reality 
( an  abstraction ). 

The more 
complex 
the model, 
the greater 
uncertainty 
becomes.

Trying to model complex 
things goes well beyond 
looking at variables that
we can actually measure.

Assumptions have 
to be made, and they 
are inherently subjective,
rendering model outputs
relatively useless 
as forecasting tools. 

That’s why most
computer modelers 
talk about “projections”
rather than “predictions.” 



Estimated mortality 
from COVID-19 shifted 
massively over time, 
as new data came in.

A new public insight 
into the limitations 
of computer modeling 
will hopefully spread 
from COVID-19 science,
to climate change science.