This paper contains one of the most important messages of how we should act when scientists argue for different strategies. Scientists need to debate and explore different strategies. But we need to act on precaution. This is VERY DIFFERENT from how scientists are trained 1/

For example, in evidenced based medicine, we want to be really sure an intervention works. This approach has great benefits for standard interventions of cancer, diabetes or other non-communicable diseases, and for re-occurring infectious diseases. 2/

If one faces a disease that is new and rapidly spreading, we cannot wait for randomised control trials. We rely on the information we have. When we debated weather we should close schools there were critics that argued about the huge economic costs and that 3/

kids may not be very contagious, so we do not need to close schools. These expert may be right (in the sense of being more than 50% likely right), but even if it was good to close schools because we were not sure 4/

the same is true for masks. No one I know claims they know 100% that masks work. But lots of evidence indicates there is a good level of protection, especially in the infected wears masks. And we KNOW THAT WE DO NOT KNOW who is infected 5/

the case against masks is very weak. It relies on behavioral assumptions, such as, people would feel a false sense of security. Actually we know this is not true in Asia, so it relies on behavioral differences of Asians and Westerners 6/

Hence in the absence of very clear evidence or counter evidence, we should all wear masks by the precautionary principle. The same was basically applied to the decision to sclose schools. I am not aware of any randomised control trial that showed school closure works. 7/

Masks costs much less than school closures, thus debate about masks is really not necessary. Many countries now recognise this. this paper will hopefully convince the remaining countries:
https://www.bmj.com/content/369/bmj.m1435
PS: people like @nntaleb saying this since start of outbreak8/8