It is interesting to compare the @IMHE_UW model with the https://gleamproject.org/covid-19#  model. For example, both predict that with strong social distancing, around 50k US deaths by April 30. However, the gleam model doesn't go past this, which I appreciate. 1/ https://twitter.com/CT_Bergstrom/status/1250304069119275009

The potential errors propagate quickly with longer term forecasts because there are so many unknowns. This is where the IHME model makes a lot of underlying assumptions about how these unknowns will behave. I think it is better to keep the predictions more short term. 2/

I don't know much about epidemiology, but I do know a lot of math. And one thing that is clear is that if the message to the public is based on math that doesn't necessarily model the problem correctly, there is a good chance that poor decisions will result 3/

If we relax our social distancing policies without a coherent federal plan to manage a response to new covid-19 infections, and if we don't have reliable data about total infections and how immunity works, it seems reasonable to me that the exponential growth will start again 4/

And then we're back where we started, having lost a lot of the benefits gained by our difficult work so far. So, I really think it's important to avoid giving long term predictions based on models that have many underlying assumptions, even though the public wants "the answer" 5/

I'm deeply concerned that there public at large and many public policy makers don't have enough math literacy to quickly understand the subtleties differentiating these models. As a mathematician, I've put in a lot of work trying to understand these for myself. 6/

So it seems to me that the more responsible public action is to make available shorter term predictions than many people desire, and try to explain what the complications are in medium to long term that make accurate predictions difficult. 7/7