Firstly, news reporters with little understanding of the subject matter try to explain it to the masses, and the results are disastrous. (Not to mention, the drive for academia to publish papers that prove the hypothesis.) Little needs to be said here.
The second is that I frequently see my peers latch onto any PDF they can download that supports their position. "Look, look! The data shows I'm right! And this man with a PhD supports my conclusion! (Of course I haven't read the report, but that's not important. What is important is my pet issue!)" Prosecutors using sloppy lab work to get convictions and computer algorithms deciding bail bonds are just end results.
We need to put the brakes on the empiricism train. It's useful, but only to a certain point. Reason eventually needs to take over. If your argument lives and dies by the next PDF or Machine Learning model, you need to find a new argument, or bolster your argument with reason.
(Meta argument: I'm reading a report about a paper someone published, and latching onto the report because it supports my conclusion that people will believe anything a "scientist" says! I'm victim to my own problems!)