> Another way of looking at is this: every year, there are plenty of people who are acquitted of crimes they were prosecuted for. Can you say all those people turned out to have committed the crimes they were accused of, simply because the people in charge of finding criminals found them to be extremely suspicious?

Given that criminals tend to have long histories of crimes and that only a fraction of crimes are caught, then yes: knowing someone faced charges ought to massively raise your estimate that they have committed past crimes which they were not caught for. Of course there are many innocent people who have been charged, but crank your Bayes theorem here...

The reason we have a very high bar like 'innocent until proven guilty' is not because this is any sort of information-theoretic or statistically-optimal criteria which balances some chosen ratios of innocent:guilty convicted, but as a check-and-balance on the power of government and a mechanism against tyranny and abuse of the laws.

> But putting that aside, the null hypothesis is that there is no statistically significant relationship between being Chinese-American and likelihood of being a spy - the burden is on you, not me.

The burden here is on any kibbitzer who wants to engage in hindsight and accuse anyone of racism for suggesting that maybe Chinese scientists are not precisely as trustworthy as natives with regard to national security.