[alpacaillama]

Would you care to post the same stats but updated for 2020 then? Also i see that this is statement is also copy pasted on your twitter so I have a hard time believing you actually do read this feedback and didn’t just post This comment as damage control.

[guy_ro]

Hi! I am really here, not a bot :)

If we're talking metrics, the 64% stat cited in the article turns out not to be a good way to measure impact of recommendations on an extremist group. We internally think about things like prevalence of bad recommendations. More on prevalence here - https://about.fb.com/news/2019/05/measuring-prevalence/. I don't have a metric I can share on recommendations specifically but you can see the areas we've shared it for so far here: https://transparency.facebook.com/community-standards-enforc...

[sudoscript]

You deny the statistic in the article. Then you use corporate speak like "we internally think about things..." Then you link to a blog post titled "Measuring Prevalence", which doesn't have a single data point with a number in it. Not to mention that it's another mess of CorpSpeak that sounds like a PR committee wrote it, not a human being.

The problem is not your intentions. It's clear that you actually want to engage people, given that you're replying actively on this board, even to impolite comments. The problem is the style of communication. It sounds constructed, artificial, developed through a process, like you and a bunch of people are sitting there drafting replies and tweaking until you have something that won't come back later and bite you in the butt. The problem with that strategy is that you end up saying very little of meaning at all.

So maybe to make things more concrete - is there a statistic on measuring the impact of recommendations that you do agree with that you can share?