From a Bayesian perspective, this story illustrates how ridiculous it is to target Muslims for additional surveillance, or to stop accepting refugees from Syria

From a Bayesian perspective, this story illustrates how ridiculous it is to target Muslims for additional surveillance, or to stop accepting refugees from Syria. Let's say there are 500,000 people with the ability and intent to commit a terrorist act. This includes 40,000 Daesh, 10,000 Boko Haram, the military wings of Hezbollah and Hamas, Maoist rebels and Hindu nationalists in India, Basque separatists, FARC in Colombia, neo-Nazis and other violent white supremacists, etc. So the Bayesian prior is something like 5e5 / 7e9 = 0.00007. Even if 70% of terrorists are Islamic, with 1.6e9 Muslims, knowing someone is Muslim only changes the posterior to about 0.0002. Since there are many non-Islamic terrorists, this is probably overstated. If the article is right that engineers are overrepresented among terrorists by a factor of nine, then law enforcement would be more justified in targeting engineers for investigation than they would be in targeting Muslims generally. That's obviously ridiculous. The posterior is so low for both engineers and Muslims that it would be a huge waste of time with an astronomical false-positive rate. Just looking at the math shows that these broad generalizations are useless.