It also involves a lot of "eyeballing", particularly the part where the author looks at the graph and notices a weird pattern. That's very hard to do automatically.
When you run a worldwide network that has significant potential for abuse and spreading misinformation I would consider it to be your duty to monitor it and guard against these kinds of things. If it costs too much then maybe you shouldn't be in business or should change your business model so that you can afford to clean up the damage your business is doing to society.
The author's analysis that this is more sinister than simply a promotional campaign is, of course, correct: the coopting of the genocide hashtag to promote the government and the use of fake, international-sounding accounts are both skeezy as all hell. But to get to that conclusion, you have a professional journalist doing a level of post-automation analysis that's too complex to operationalize at scale. It requires too much critical thinking and discretion, and throwing a million human reviewers at it will just result in the kind of capricious false positives that are also railed against on HN and elsewhere (perhaps rightfully), with an undoubtedly high amount of false negatives too.
There are constant ignorant-layman calls for magical tech moderation that only touches bad targets and never steps on the good little guy, but none of the people making these blithe demands ever stop and think about whether this is possible, beyond "they're rich and smart, they definitely know how to do it".
So you do need people somewhere in the loop to handle these things.
What would be nice though would be if social media offered rewards for identifying such abuse, after all - they happily pay bug bounties, why not abuse bounties! That would get them lots of zero-hour/non-contract workers doing their job for them and big companies love that it seems and to be able to do that in a PR way that has positives. Well, it works for BUG bounties, why not incentivise the populus to police the populus against abuse as well.
...
> if social media offered rewards for identifying such abuse
couldn't this also be gamed for false positives, and create an incentive for groups to form smear-for-profit entities which try to spin small communities as 'bad' to the benefit of others?
But let us not forget, things like posting IP and client and other such details that may prove useful in identifying such abuse, may well elude public investigation. So it would be these niche area's in which Twitter could then focus upon.
However, as an example of a slightly comparable problem - wiki editing and the weighted credibility of editors over time, we may well end up creating a hidden social media based upon a small subset of society and perhaps yielding a bias in some direction or another in judgment of what is abuse.
But with the ways social media churns upon spam/botnets, it would place a huge load upon people self filtering.
It is not unreasonable to take the top 5k accounts/hash tags and manually evaluate the top replies and retweets. Human evaluation can be very streamlined with clean evaluation rules, and these human evaluations get fed back to machine learning models.
And after all that, Apple's recent products have been somewhat lackluster. Catalina is currently a dumpster-fire; a friend installed it and had his machine just sit frozen for a while along with a huge number of other problems. Keyboard design, too. Doesn't sound to perfect to me.
Twitter’s failures on the other hand affect everyone - misinformation spread through their platform can and does affect society at large regardless of whether you’re using Twitter yourself.
This happens much more often than you might think, there are a quite a few examples of this effect.
https://grrrgraphics.com/wp-content/uploads/2018/07/jack_dor... https://pbs.twimg.com/media/DTmD9bmVMAA9M19.jpg https://static.thefederalistpapers.org/wp-content/uploads/20...