IHME Model flip flips on predictions -- How do we go about interpreting the revisions?

It appears as though the IHME model has in the past week adjusted its prediction of deaths first upward, then downward below the previous estimate. Data updating is changing the projected outcome. This is as it should be, but it highlights the problem of relying on models of an unfamiliar virus to drive policy.

And if I understand correctly - emphasis on IF - the earlier higher projections were based on assumptions that strong social distancing measures were in place. This suggests a flaw in the model's assumptions about the effects of social distancing.

We are seeing frequent revisions downward. That doesn't mean Covid-19 is not actually a big deal. It is, as far as I can tell. But it is a typical pattern in politics to set initial claims at a level that motivates action, then update those claims sequentially to become more realistic, one we have committed to a course of action.

We see this with megaprojects, which are consistently underpriced initially. As San Francisco politician Willie Brown said of one megaproject, the public should have known the original cost wasn't really what it would cost; if they had told people the real cost up-front the public would never have approved the project.

Now that the lower IMHE projections are based on the assumption that current social distancing policies and behaviours continue through May. If they don't continue, we could still have more deaths than the model currently projects.

I'm skeptical we can maintain these policies and behaviours for that long. So I really hope their model is too pessimistic on that.

I've been asked multiple times what motive the experts would have for being less than wholly honest. As a political observer it seems quite honest to me. Experts are human. They are almost never wholly and perfectly objective. The experts here saw danger ahead and wanted to spur politicians and the public into action. It ain't rocket science, folks.