Hacker News Reader

Comment by johnc1231 | Hacker News Reader

I think Twitter's biggest problem with fake accounts is not that they are hard to identify, but that if they do identify them and shut them down, it'll hurt their "number of active users" stats

View on HN

I suspect their big problem is more that their site is too big for the moderation staff they have at the moment. Identifying bots is easy for people sure, but it's hard to automate with AI and hiring people to proactively hunt out and shut down bots is expensive.

Of course, the fact doing it too well hurts the stats doesn't help either.

> it's hard to automate with AI

I disagree. At least it shouldn't be that hard for a company with Twitter's level of engineering and data science sophistication, and yes, I do realize that most of their best technical talent is long gone.

> it’s hard to automate with AI

While nothing is a panacea, there’s little evidence that Twitter is doing even the bare minimum.

Their Perverse Incentive is probably to shut down fake accounts faster than they are created. Such that they are 90% effective in, say, 5 years so that the active users takes a haircut instead of a big drop.

I used to work for a company where we were explicitly told not to remove the bots as active visits because it would hurt the figures we were giving our customers. For The PM team having so much traffic was a blessing regardless where it came from. Not same problem as fake accounts but still similar.

> Not same problem as fake accounts but still similar.

Actually, I think it's almost entirely the same problem as fake accounts, from Twitter's perspective at least. It's all a measure of usage and demand by customers, and when people are used and investing/involved/compensated for that fake increase, any major drop rocks the boat.

Reply on news.ycombinator.com