You’d think they’d have learned that a general list of “nono” strings would be a bad solution to the problem of identifying problematic language, especially on a multilingual platform. We ran into an issue identifying action statements where it said “managing director” was an action. Which it kind of is grammatically, but not in the broader context we worked in.

This is what gets me about these “citizen scientist” types. Don’t you think the scientists carefully crafting a rigorous analysis that needs to pass extensive peer review know enough to make sure that “Suicide Squad” isn’t captured by what they’re searching for?

How do you think that the experts are dumb enough to miss an exceedingly obvious confound instead of correcting for it or at least being aware of it? It’s like the anti-climate change people looking at a paper and going “did they account for volcanic activity” as if that’s some huge missing link that was likely overlooked.