My experience was:

a) The med community was *terrible* on masks as of late March -- people I spoke with insisted, "We can't recommend something that wouldn't be approved by the FDA."

b) Epi community was more reasonable, but many mistakenly touted "mixed evidence." https://twitter.com/aetiology/status/1277338668630507523

The med community was hyperfocused on getting PPE for medical professionals, obviously a critically important goal. But because of this, they were opposed to calling for widespread *cloth mask use* that likely would have saved tens of thousands of lives.

The epi community was more split. Some (like my co-authors Sten and Albert) were 100% on board with calling for universal cloth mask adoption. But others were much more ambivalent, I think for two main reasons.

First, they pointed out that existing RCTs did not find benefits of masks for SARS-like viruses. For example, this article: https://www.bmj.com/content/350/bmj.h694 reviewing the RCTs that had been conducted in community settings.

But there were three serious problems with this reasoning: 1) NONE of the RCTs were designed or powered to assess whether masks prevent transmission of the virus. This can't be done by comparing outcomes for people with and w/o masks.

2) Many of the studies did find an effect for masks & hand sanitizer, and given other effect size estimates for hand sanitizer, this arguably suggested that masks were effective (although only weakly, since different samples & low power).

3) The low compliance rates meant that the studies were generally very underpowered. However, they did not properly adjust their standard errors for "mask effectiveness" given this low compliance rate, they only reported the (insignificant) intent to treat estimates.

These studies were NOT mixed evidence. They provided little evidence of any kind about masks, and if anything, suggested that masks probably work. The compelling evidence was not from these studies, but from lab studies which consistently showed masks block viral emissions.

Second, some epi people argued that masks were probably good if used properly, but might backfire because they would get people to engage in riskier behavior or complicate messaging about social distancing.

Again, there was *no evidence at all for this conjecture*, which contradicts decades of experience from other safety devices like airbags and seatbelts and helmets which suggests that risk compensation is not dramatic enough to overwhelm direct effects from (even imperfect) use.

This is not a general indictment of the epi community -- many I think reached the right conclusion, but others were too slow to recommend a critically important solution even after there was clear evidence of asymptomatic or presymptomatic spread.

I see two lessons:
a) Experts in one field can go awry when they base their judgments on intuitions about stuff they are not expert about (e.g. risk compensation).
b) "p = .05 in an RCT or bust" is terrible epistemology -- underpowered null effects are not evidence of no effect.