Hacker News new | comments | ask | show | jobs | submit login
Battling Fake Accounts, Twitter to Slash Millions of Followers (nytimes.com)
176 points by nthitz 7 months ago | hide | past | web | favorite | 112 comments

Kudos to twitter on this. This is one of those moves that harms them in the eyes of short term investors, but is The Right Thing To Do.

I've always been astounded that nobody seems to notice that Facebook continually announces record users, record MAUs and then usually a week or two after their quarterly numbers announces that they have removing millions of fake accounts. Always after they've 'beaten' on those metrics.

Facebook these days, for me, is 90% dealing with fake friend requests from people that are either clones of real profiles, or just complete garbage. It's part of my dislike of their platform, but they're more afraid of bad metrics than genuinely delighting their customer base.

It becomes a negative feedback loop too. We used to spend significantly on Facebook for several businesses I'm involved in, but now that most clicks are garbage/fake, we don't.

I know twitter gets a bunch of hate for not doing more, but realistically they are actually leading the way. I'm impressed with how they're operating more and more.

As a twitter user, it doesn't matter to me. I don't see these bots & fake users. What annoys me with all these platform is that that they choose what to show me. I wish to see all posts/tweets in the exact order it was created. That's it, don't decide who's I should see. Every once in a while, I'm shocked to come across someone I was following and havent' seen their tweet only to check their profile and see they have been tweeting tons! Meanwhile, I see tons of tweets from people I don't follow only because someone I followed retweeted it. What I've done is to unfollow a good amount of folks to reduce the noise.

> I see tons of tweets from people I don't follow only because someone I followed retweeted it

This is the original, only purpose of retweets, going back to before it was officially supported and was done manually.

Showing tweets which have been liked by accounts you follow is a much worse behavior of the app. I like indiscriminately but tweet and retweet very seldomly. There's no feature to save and show appreciation for a tweet without boosting it to my followers.

Favorites are useful as an "ack bit" too: indicating to someone that you've read their reply to you when you don't have anything else to add to the thread. I am more hesitant to use them that way these days.

This. The number of people I follow are in the low double digits. I follow them because they are industry leaders and it is how I keep up on the latest news and new technology in my field. The new algorithms are showing me tweets from days ago or just skipping tweets completely. There should be an option to see in order or by algorithms.

It's the way social media is going.

From google news to facebook to twitter to youtube, instead of seeing what you want to see, now you see what they want you to see.

Social media is now parlaying its giant audience into social engineering. Buckle up because it's going to be a rough ride.

I don't see these bots & fake users.

Pop up detail view of a popular tweet from a popular-in-tech person sometime.

You'll probably see, right below it, a reply from an account with identical profile picture and display name to the popular person, but different username, pretending to continue the thread and announcing a cryptocurrency giveaway where you "just" have to send some amount of ETH to their address to "register" or "verify" yourself.

Pretty much any popular Elon Musk tweet will have some kind of "giveaway" impersonation scammer auto-replying, for example. For a while some of the popular cryptocurrency people actually changed their display names to "I'm not giving away ETH", or similar, to try to disrupt the scammers.

Exactly this.

My other beef with twitter is a more sophisticated way to filter my timeline. I mean something that would make sense to a programmer, not the typical user friendly useless thing.

Yes I haven’t become a bigger Twitter user because it’s a huge time suck to actually get to the few topics/groups I care about. Medium also needs to do similar.

Agreed. Basically, it is akin to having a second fulltime job. Except you don’t get paid anything.

yes! I really wish it had this. I have several interests. Sometimes I want to see the news, sometimes read about tech, sometimes medicine, whatever. It all just runs together and makes me less productive instead of more...

List work well fo that purpose. Many clients also allow imeline filtering, or saved searches (with Boolean queries, and including flags like „people I follow“

If only there was an open to mass 'unfollow' I could have used it. Instead you have to manually do it individually while confirming 'unfollow'

I hardly use it but i checked my account last evening skimmed through my feed and saw nothing productive. What caught my attention though was a Forbes tweet projecting one of the Kardashians becoming a billionaire because of some cosmetic empire?

I dont even follow anything fashionable except Thomas H. Ptacek and Paul Graham.

I instantly deactivated my account pending deletion. Twitter is as much of a cesspit as Facebook.

Since I've thought the same thing 'I've followed these people but never see their tweets'

I googled around as to whether I could change the feed filter/ algorithm

Looks like if you go Settings/ Content Preferences/ there is a box for 'Show Best Tweets First' that you can uncheck,

which looks like it might shut off their recommendation filter, and then you can just get a stream of whatever everyone you follow posts

(or at least that's what I hope it is, I've had it set on that setting for all of 2 minutes, still seeing how it works)

You can separately unfollow retweets from accounts that do that too much.

You sure can, but you can't turn off seeing what people you follow liked (unless you use something like Tweetdeck or an app that doesn't support this 'feature'). I find that worse than retweets because a lot of people apparently use the 'like' button as a sort of read receipt and 'like' everything that shows up in their feed.

There is actually a foolproof way to shut off Liked tweets (and retweets) forever. You can only do it on desktop - but go to Twitter/Settings/Muted Words. In there, make sure "From people you don't follow" is selected along with "Timeline", then individually add the list of common stopwords to your muted words. I listed the ones I use below, you can add more but I've not seen a retweet or liked tweet since doing this. muted: [a, about, an, and, are, as, at, be, by, for, from, has, he, her, him, his, how, i, in, is, it, its, of, on, one, or, she, that, the, to, was, were, what, where, when, will, with]

Clever! But if you have selected "Timeline", aren't you muting both tweets and retweets with your stop words? Twitter's UI makes the "From people you don't follow" option look like it only applies to "Notifications", not "Timeline".

Alternately, you can control whether you see what individual people you follow are retweeting with "Turn off retweets": https://twitter.com/following

yes it will mute retweets too. the desktop ui allows you to pick only the people that you do not follow - the interface and options are different on mobile, just fyi. they don't make it very clear either way.

>/following yea really wish you could do that with likes too!

> but you can't turn off seeing what people you follow like

Yes you can, not through an official setting unfortunately. But as TylerH said you just say "I don't like this tweet" on a few of the tweets that you see because "person X liked this", then refrsh your feed and they're all gone.

Maybe after 6 months-ish they come back, but you notice it really fast since suddenly your twitter feed sucks. Just do the "I don't like this tweet" trick again.

The only 'likes' I see now is if multiple people I follow liked a tweet from someone I also follow. Usually those are good quality.

I actually deal with this by marking every "liked" tweet shown to me as "I don't like this Tweet" whenever it happens, and then after each time I tweet at Twitter and TwitterSupport to stop showing it to me ever. Works well for about 3 or 4 months until I get another batch of 10 or so "liked" tweets.

They can't control the narrative if they show every tweet.

"... I wish to see all posts/tweets in the exact order it was created ..."

Oh my, that definitely should be filed under: "Be Careful What You Wish For"

I suspect such a change as that would end up annoying the living you know what out of most every user on Twitter. They should let YOU configure YOUR OWN account to see everything. But I'd be really hesitant to do that for every account on Twitter by default. That just would not end well.

You can make Twitter display all tweets in chronological order, they just call it a "list". Instead of following people, add them to a list. When you view the list you'll see all the tweets from people on that list, in chronological order. And as a bonus, no ads!

I'm OK with this being OPT IN, hidden behind three layers of warnings. I just want it to be an option.

It is an option, a terribly obscure one, if you use Twitter's search interface to read your tweets instead of your home timeline.

I have this URL bookmarked as "Twitter classic":


> Kudos to twitter

Kudos? I wouldn't hand over a kudos to twitter.

The article triggering this bot ban wave [1] was written by the Times and included behavior analysis that Twitter easily could (and may have) done on its own years ago.

Facebook's actions and marketer value has nothing to do with Twitter's own lack of action to prevent infiltration and manipulation of its platform.

Twitter's action on stopping spam, bots and hate speech at best can be described as slow. I believe Jack Dorsey has repeatedly said the company hasn't / doesn't do enough.

[1] https://www.nytimes.com/interactive/2018/01/27/technology/so...

He keeps on saying they don't do enough, the times is able to do this despite not even having access to raw internal data

I feel like Twitter as an organization doesn't actually care about the problem, just about their image.

Why else would they talk about the hate speech stuff and continuously elevate groups like The Proud Boys through their verification process (that also affects their suggestion algorithms). And we shouldn't forget that many Gamergate people are still active on Twitter and often use their accounts to direct harassment to people on a regular basis!

The even more cynical view is that there's a group of employees within Twitter actively protecting these kinds of users

That's the problem with companies once they go public, and I'd argue once they accept any kind of investment; the new owners have the company by the balls and expect Results, expect the stock price to rise. Twitter's artificially maintained that stock price by not banning their users, because it seems user quantity and growth are far more important metrics that influence stock price than the quality of users.

Twitter cares about its bottom line.

If otherwise, they'd have suspended Trump's account for ToS violations long ago.

> We used to spend significantly on Facebook for several businesses I'm involved in, but now that most clicks are garbage/fake, we don't.

It seems to me that the rational reason to stop an ad campaign is because it stopped enough making money.

It's not fake clicks that should be stopping you -- it should be a lack of ROI.

I wonder how many people spend huge sums of money on Facebook without tracking their return on investment. I seriously think that is the source of many of these complaints about Facebook's ad program.

I spend a few thousand a month of Facebook. I have no idea if there are fake clicks or not. But, it's not a metric that seems to matter in my business.

While it’s a good point that the “quality” of click should be a proxy (i.e. you care that RoI might be low because of bot visits, not about the bot visits in themselves) rather than an actual measurable, is it not reasonable to assume that if “most clicks are garbage/fake”, they’re not making any money from those and therefore are unlikely to make “enough money”?

I guess that assumes similar cost per click and that your definition of “enough” accounts for a sufficiently low conversion rate.

On the other hand, Facebook (and their shareholders) probably cares that the PPC is depressed by the fraction of non-converting users.

I think you're overstating the frequency of fake friend requests on Facebook. I'm fairly active on there and usually receive less than one per week. (It was worse a couple years ago.)

It may be a 'type' thing. Every time I look at my neglected Facebook account I have anywhere from 6 to maybe 15 friends requests from accounts which have basically no timeline and a "sexy" picture of a 20 something female. If those accounts do have any friends it seems they are all white males over 50 years of age.

I once tried to give Facebook another go and created an account again. It seems like my email was in the meantime leaked to some gambling spam list because 90% of my friend suggestions were fake profiles from around the world created purely to log into a gambling mobile game and (I assume) farm coins/items/etc.

The accounts were obviously fake - they weren’t even trying to hide they were fake, and I actually spent maybe 20 minutes reporting 100 of them, but not only was there an endless supply to replace those I blocked, but the reports didn’t do anything and the accounts were still there.

Instagram has a similar issue as well. It's brain-dead to buy fake followers, have them simulate interaction with your posts to push you higher in the algorithm ranking, and bots are the majority of the user-base from what I can tell. And they have no intention whatsoever of fixing that.

> This is one of those moves that harms them in the eyes of short term investors, but is The Right Thing To Do.

[Cynic's hat on] And they can only do this now, having shown the ability to turn a profit.

Up until now, Twitter has been forced to operate in "growth for its own sake" mode. Investors keep myopically staring at user numbers, confident that there's a way to turn the faucet on, consistently and repeatedly tapping the userbase for pennies per person. The bot accounts don't make them money, so getting rid of them are likely to actually improve the bottom line.

But the sad fact is that until recently Twitter have been financially disincentivised from cracking down on fake accounts.

> Facebook these days, for me, is 90% dealing with fake friend requests from people that are either clones of real profiles, or just complete garbage.

It's possible to block friends requests from anyone except friends of friends.

Have followers on Facebook and filter your friends requests. That is what I do.

My mother and father have each had 'fake' clone profiles of them created a few times.

Occasionally I get friend requests from some random dolled up woman, perhaps once a month.

... that's it.

I wonder what causes one person to get a large number of bogus friend requests and others not. I'm guessing it has to do with who your friends are, and how many?

I'm around 1000 FB friends +/-, but the vast majority of them have been on Facebook for a VERY long time, so perhaps they're less likely to be friend scam accounts?

You may already have a fake friend in your Facebook friend list. I think your account is marked as a target because of it. It's not something most accounts experience.

We’ll see. Twitter has a history of being overinclusive of conservatives in assessing bot accounts. If this purge is just more of the same, they don’t deserve any congratulations.

Oh my absolute fuck. Every single Elon Musk tweet is followed up by an account with an identical image, and a one-character-off twitter account pretending to continue the conversation with a 'surprise'. It's always a link to some crypto BS. How in the hell Twitter can't auto-block that, particularly when the source account is a verified one, is behind my comprehension.

The problem is so much worse than this now. There are dozens of fake accounts under every one of his tweets and it doesn't even seem like these accounts are trying anymore. Their names are sometimes random strings of letters, and they just change their profile picture to a screenshotted version of Musks. Oh and once you report one of these accounts it can take weeks for Twitter to do anything

I also find it so strange that Twitter hasn't done anything to specifically limit these accounts. They could even naively block normal users from having a verified user's profile picture and this problem would virtually not exist.

> They could even naively block normal users from having a verified user's profile picture and this problem would virtually not exist.

I'm all for Twitter doing more, especially as they seemingly do nothing at all to combat fake accounts. However I'm sure these spammers could just adjust a few pixels on the profile image and a naive block wouldn't do anything.

Yeah I agree with that. Maybe they'll need a more sophisticated solution that can check for visual similarity. But It just feels like they haven't even acknowledged this as a problem, so I would be happy with any countermeasures right now.

I noticed a TON of Binance related spam when searching for specific ticker symbols (Nasdaq/NYSE stocks, not coins). It's since gone away from what I search, so they're doing something right, but spam is a tough problem.

Finally someone has some perspective on things.

Odd that they counted in the first place. Maybe this is part of a phased roll out to identify first (refine the models), then take punitive action?

The cynic in me says they counted in the first place because it helped everyone's ego (and wallet)

"Most of the time, according to Twitter, the locked accounts are not included in the monthly active user count it reports to investors each quarter, a critical Wall Street metric for social media companies. But the locked accounts were nevertheless allowed to inflate the follower counts of a large swath of users.

That choice helped propel a large market in fake followers. Dozens of websites openly sell followers and engagement on Twitter, as well as on YouTube, Instagram and other platforms. The Times revealed that one company, Devumi, sold over 200 million Twitter followers, drawing on an estimated stock of at least 3.5 million automated accounts, each sold many times over."

> The cynic in me says they counted in the first place because it helped everyone's ego (and wallet)

I don't think you're being cynical. They counted in the first place because it helped their _metrics_ – Twitter is in the business of selling advertising and made the decisions they did in service of their advertising metrics.

I remember skimming their quarterly report a year or two ago and being shocked that they focused so much on user acquisition as a core KPI given it was clear how goosed those numbers were. I was expecting slightly more complex CLV type numbers but instead they were fawning over a glorified "select count(*) from users".

There's really only one solution to Twitter's problem, and anything short of it is just nonsense placating. They have to open up profile verification to all applicants. It would fix the platform overnight. You could just immediately ignore anyone who's not verified, yet it would still be optional for those who don't care.

They started to do this but then paused all verifications because bots were getting verified.

I saw one of those Ethereum scambot dopplegangers on a high profile tweet that was verified. Really looked awful and definitely caused more damage since people are trained to trust it.

Any hacker that manages to get the credentials of any verified user can just change that account's name and picture to that of a famous person for easy impersonation.

Just the fact that you can get a verified account and change the name to anything without the need to get verified again is a massive loophole.

Yes. Now that the prestige value of the blue checkmark has already been torpedoed by their various verification fiascos, it makes perfect sense to make the blue checkmark available to everyone because it would also solve a lot of their problems in the advertising department.

Right now clicks on Twitter ads are worth very little because most of the clicks come from fake accounts. But imagine how much clients would pay for "verified clicks" -- clicks from verified accounts.

IMO they should make their ad sales a tier-based cost-per-click model with clicks from verified accounts selling for higher amounts than clicks from non-verified accounts. If theyre smart they'll open up a third tier above "verified" for the twitter-endorsed superstars which will have prestige value again (now that the blue checkmark has none). Maybe a green checkmark would work. Naturally they could sell an ad click from a green checkmark account for even more than an ad click from a blue-checkmark regular ol' verified account.

There's really only one solution to Twitter's problem, and anything short of it is just nonsense placating. They have to open up profile verificatio

There’s another solution: the platform is free if you have less than n followers or follow less than n users. Say n is 15 for example. Above that you pay for every tweet. On a sliding scale, say the first tweet/day is 5c, the 10th is $1. They could even ditch scammy advertising - brands who wanted to “engage” would simply pay directly to tweet.

I'm sorry but that is just a horrendous idea unless N is a very high number

Why do you think so? It would solve all Twitter’s financial and social/cultural problems practically overnight

Twitter needs its users infinitely more than Twitter's users need Twitter. Charging anyone to use the service at all is counter-productive.

I'm curious if this is a real crack down or they are just trying to act like it.

All of the big social networks have tons of bot activity to the point that most likely the majority of their requests are from bots. If they actually stopped all bots activity it would tremendously hurt their user counts and thus their valuations.

I'm curious how many "suspicious" accounts will be real people simply expressing their views. Twitter does openly admit to censorship [0], so it's not exactly a wild thought.

[0] https://www.lgraham.senate.gov/public/_cache/files/4766f54d-...

Human logs on ,sees suspension notice, goes through unsuspension process.

Bot dies.

Those humans are usually very bot-like retweet mills anyway, in my experience. Thou complain'st too much.


Human logs in, sees Twitter demanding personal information like a phone number or worse, goes away.

As a person who buys ads on Twitter and Facebook, I really need them to step this up. It is killing customer acquisition costs when the majority of clicks come from fake accounts.

Surely Twitter would raise their price per click when they increase the average quick quality. Caring about the value per click is ridiculous for a customer -- you should be willing pay whatever rate you think is is worthwhile for the aggregate value of the traffic you see, a number determined by experimentation.

For sure, but I can't control the quality of the click. I'd be willing to pay a lot more per click if there was any realness to it.

We stopped our (tiny) $1500 per month Facebook campaigns for exactly that reason. So scammy: we pay real money for a bunch of bots to “engage.” Until robots start buying our product, I won’t be buying ads where bots are the largest source of engagement.

I regularly see comments like this about Facebook's ad program. Personally, it strikes me that there was no tracking of ROI in the first place -- and if you can't reliably track ROI, why were you spending $1500?

If you're spending $1500, but getting great ROI, then fake engagement would be no reason to stop the ads.

People are tracking ROI, and constantly testing new targeting parameters to keep eking out a few more points of ROI. A lot of Facebook Ad spend is experimental (especially at that low of a price). There used to be a time when you could get real engagement for pennies each and build up a following of real people based on a few ads then not have to spend anymore and go off your organic growth and audience.

Now the second you stop the ad spend your engagement drops to nothing.

What's wrong with even 99% robot engagement is 1% human engagement pays the bills?

People don't normally knowingly pay $1 for a penny's worth of value. Think of it like a bucket of apples. If I charge you a price based on 100 real apples and give you a bucket that only has one real apple in it that's basically fraud. Doesn't matter if that one apple is delicious.

However, since it's auction based, if you were getting real clicks, the cost would be significantly higher than $1. I'm not sure how it's a penny's worth of value, given that the price includes the market's best guess as to the value of the mix of valid/fraudulent traffic.

It'll be fun seeing which accounts lose the most.

I expect a lot of whining about it, particularly in the political realm.

Instagram is a space where influencers were deeply affected by a purging of bot accounts (although I can't remember the year this occurred). People made posts in droves complaining to Instagram, asking for them back, and in one case a user cried that "those were users I will never get back".

Interesting to see social media platforms finally starting to be (somewhat) more accountable to their users. Their motives are a dichotomy. On one hand, their revenue comes from advertisers, so they must be attractive in the "metrics," but by padding the metrics, they lose the users that those metrics are supposed to reflect. It seems that Twitter has finally reached an inflection point where having more engaged users is more important than just having "good numbers."

As social media matures, I'm sure we'll see where the advertiser-friendly vs user-friendly line is drawn. Reminiscent of the banner ad boom that initially made great revenue for the sites, but in the end netted negative for the sites because of the users it drove away.

These are not bots.

I've encountered many people complaining that Twitter thought they were a bot. Sometimes that was that, The End. Sometimes there was a demand for "real world" identification, like a phone number, which got refused because people like their privacy.

The common feature seems to be following or retweeting political users toward the right, and using hashtags associated with that. Do that exclusively, and twitter will assume you are a bot. The same does not seem to apply if you are on the other side of the political spectrum.

I think Twitter is well aware of this, and they consider it an intended result, but they can't just publicly admit it.

That may not be Twitter's fault exactly. Twitter is overrun with left-leaning SJW types and why is the W short for "warrior"? It's because again and again these people use whatever tactics they can get their hands on to defeat their perceived enemies i.e. anyone with conservative views.

Most likely there are people on Twitter who make it their mission to report conservative tweeters as spammers, knowing full well that it will cause Twitter to auto-harass them on their behalf. They do it in the hope they can get these accounts shut down and thus "cleanse" Twitter of wrongthink.

This sort of behaviour should be easy to stop, but frankly Twitter's spam team was never that good (just check account prices on the black market). Lots of trivial techniques were never implemented by them.

Good. If you follow any remotely popular people you'll run in to bots. You can see bots spamming replies to people even with just 80,000 followers --less than the million mark.

The reality is, that Twitter could have easily and simply fixed this issue long, long ago.

But, to do so they would have to place some power in the hands of the users. They don't want to do this because it means a loss of their control over what you see.

We already know that FB experimented with pushing people's mood and views on subjects around; Twitter is not immune to such temptation also.

Simply: assign 2 scores to every user, each out of say, 1000 (100 is not enough given Twitter's large userbase).

1 score is a category score, such as "lgbt politics" or even finer-grained than that, easily done by simply reading all of a user's tweets and using Bayesian (or some other classifier) auto-classification. This is a measure of how the user is perceived by others, inside that category.

The other score is a combination of per-region or per-country (because in general, people care about the people in their own country or region more), plus number of times a user gets a like or retweet in that same region, etc.

Then, simply let people filter based on those 2 scores. If I set score 1 at (cutoff everyone below 900) I will get top tweeters in each subject; if I set score 2 at (cutoff 200) I will let most tweets across the world reach me. etc. etc.

I wonder if bot sellers will see an increase in sales? and Twitter announces a lot of new users soon.....

The comments here are overwhelmingly misunderstanding twitter‘s action: it will not lower their MAU numbers, where these accounts had already been excluded. Twitter is now just following up by also removing spam accounts from individual accounts‘ follower count.

This has been the elephant in the room for Social since the beginning.

if you've invested in twitter, join the class action fraud lawsuit where twitter knew about the 95mil+ fake accounts (conservative estimate), cause it's 101 obvious fake accounts that any idiot could spot, and still projected growth numbers including these accounts, and used those to help raise money. classic silicon valley fraud.

How can an account be fake? An account impersonating someone else? That's very different from an account that exists solely to boost follower counts. What's the problem with people paying to boost their follower count?

There really isn't, but then again, I don't know why they'd pay at all; it's hilariously easy to get a "following" on Twitter by simply following random people. I never use Twitter, but I once tried just following random people to see what would happen and I had about 200 "followers" within a few days, all from maybe an hour or two total of randomly clicking. Now 200 is peanuts, but imagine if I had kept it up for longer! Hahaha. It's so ridiculous how easily people will just follow each other on Twitter. It means virtually nothing, with or without fake accounts or people paying to boost their count.

the amazing thing is, much of the comments in response to Donald Trump and other major figures are real. the bot problem probably has to do with fake followers, hahtag spamming, and keyword spam of sorts. Much of it is invisible to the naked eye.

Non-sarcastic reply since original one was flagged:

It's sometimes rare to see people bring up the chance of false positive identification, though two comments did in reply to another downvoted comment. In the replies to popular tweets are a mix of replies from real people, replies from bots, replies from real people behaving in the way a bot does, and replies from mixed accounts where people sometimes tweet themselves and sometimes use a tool to automate tweets. One account I saw just accused others of being bots, their twitter history was filled with such accusations. I'm pretty sure they weren't a bot though based on the other tweets, and I doubt there's a tool just for that... Maybe people need to be accused of being a bot by a non-bot (sarcastically or not) to think about the chance of false positives more often? But it's frustrating to see congratulations as if the hard problem of identifying accounts that need to be banned (with some bad bots being a special subset of that) has been solved.

Jach 7 months ago [flagged]

This sounds like something a bot / Russian troll would say.

Please don't break the HN guidelines, which ask you not to insinuate astroturfing or shillage without evidence. This is a particularly toxic variant of that poison, and it's important not to use it here.


I guess I forgot to add "/s", which I didn't think was against the guidelines (just low value).

Crazy that Twitter still sells ads and soooooo many of the accounts are bots. They have an obvious interest in inflating the total numbers.

It's a super, super simple problem. If they really cared, they'd eliminate bots entirely. Not that hard.

> It's a super, super simple problem.

No offense, but that's incredibly easy to say when you're not the one responsible for doing it.

It might turn into a game of cat-and-mouse, but Twitter just hasn't been much of a cat so far. Finding large bot networks is pretty easy according to the researchers who do it all the time. https://bahrainwatch.org/amanatech/en/investigations/we-dema... https://medium.com/@robhat/an-analysis-of-propaganda-bots-on... https://www.bbc.com/news/technology-38724082 https://link.springer.com/article/10.1007%2Fs10844-017-0494-...

No, it isn't. It's not easy at all.

The researchers who keep publishing these papers and the patsies in the media who keep repeating them are not identifying bots, in fact they can't know what their own accuracy is because they don't have any way to force verification of accounts. But it's apparent from reading their work that they're usually just identifying humans.

I used to do anti-bot detection and anti-spam work for Google, and have written a takedown of one of these media articles about Twitter bots:


The paper I studied was riddled with nonsense, fraudulent claims, claims that didn't match their own data, abuse of logic and statistics and massive political bias. It is typical for this space.

This is especially true because real bots (vs the bots that exist in the imagination of political journalists and academics) usually want to sell things. They're quite boring. They aren't engaged in "propaganda". The belief that spammers are spending lots of time and money trying to flood Twitter with political opinions isn't grounded in any sort of reality, and frequently leads to embarrassing claims, like this guy who was pegged as a "Russian bot" but who turned out to be a carpark attendant in Glasgow:


These claims about spambots should be seen in the context of a US political campaign in which the losers seemed to be struggling to understand why their preferred candidate lost. The idea that the population has been bulk brainwashed through Twitter doesn't have any basis in psychology or computer science, but is an attractive way to avoid engaging with policy issues.

Verifying humans is a solved problem.

Even assuming this is so, what's to prevent BotCo from hiring some humans to click through the process for each bot account?

Not really. You come up with a method and someone will come up with a way of factory-farming their way around it.

Please fill us in.

I share your skepticism.

CAPTCHAs are at best a game of cat-and-mouse, and are always open to relay attacks anyway.

You can still verify in-person, and even check documents, but no web company would even consider doing that, even for something like an optional 'fully verified account' tag.

Even something like checking a phone number just means that you're limiting the number of legitimate "bots" or alternate accounts people can have.

Indeed, that's a good way to raise the bar.

For small websites, even a completely trivial email check is enough to keep out most spammers.

Yeah, that sounds practical for every start-up that needs to do human validation.

It's not a simple problem. It's a complex economic calculation with at least a few variables where the false positive rate directly impacts the bottom line as well as customer happiness, and the battleground is constantly changing as spammers get more advanced over time.

Force real human verification for real accounts, and then create a bot/computer/AI account for the bots that people actually use. Human verification is a solved problem for the most part.

Eliminating all bots is not desirable. Many of them are useful and interesting.

Eliminating the bad ones with few false positives is hard.

Bots shouldn’t be allowed to follow people.

I'm surprised to see comments like this on HN.

Technically speaking, I'm not sure it's even possible to tell for sure which accounts are bots and which aren't. How can they tell?

There are plenty of ways, Javascript is an amazing thing. For instance, anyone running this site can see that I typed this in character-by-character, and even had to hit backspace a few times. Dumping all of the text in at once would be suspicious. Not a clear indicator of a bot, but one indicator at least. I also moved my mouse and just expanded the text box so I could see my thoughts. Not something a bot would normally do. I'm also coming from an IP address that only ever posts to one account, not to multiple. My browser fingerprint only posts to one account. My browser fingerprint shows me on a Macbook using Chrome, and my cookies indicate I have a web browsing history. I upvote. I downvote. I post something, then engage in a follow-up discussion later on. These follow ups are upvoted by other accounts that match all or most of the above criteria.

My day job is information security, specifically working with a SIEM to correlate many diverse logs from many diverse systems and figure out what really happened using many pieces of individually benign data. None of these things are themselves indicators of bots, but the more you start to trip these rules, the more bot-like your behavior becomes. Eventually it paints a picture that shows no human could reasonably be behind an account that routinely posts two or more tweets at the same time, never engages in follow-ups, is only followed/liked by other suspicious accounts, and has a user agent of Python 3.7 coming from a source IP on aws.amazon.ru. You show them a captcha and if they fail or bail, you've got 'em.

You could just fudge the numbers a little bit and have the system say "~X human, ~Y bot followers".


Applications are open for YC Summer 2019

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact