But no matter how many times we marked them “false,” stories would keep resurfacing with nothing more than a word or two changed. This happened often enough to make it clear that our efforts weren’t really helping, and that we were being directed toward a certain type of story — and, we presumed, away from others.

What were the algorithmic criteria that generated the lists of articles for us to check? We never knew, and no one ever told us.

There was a pattern to these repeat stories though: they were almost all “junk” news, not the highly corrosive stuff that should have taken priority. We’d be asked to check if a story about a woman who was arrested for leaving her children in the car for hours while she ate at a buffet was true; meanwhile a flood of anti-semitic false George Soros stories never showed up on the list. I could never figure it out why, but perhaps it was a feature, not a bug.