Why are people mean on Twitter? - Smarter Every Day 214
- You've probably heard a lot of talk lately about bots on Twitter or even foreign involvement in our political process. For example, the president of the United States publicly thanked a Twitter account which we now know was run by malicious actors located somewhere in Russia. Now, depending on what your political views are you might be inclined to roll your eyes and make some kind of joke about the president being tricked, but this is far more than that. This reply is totally something I would do. If an account with over 100,000 followers tweeted something nice to me I think an appropriate response would be thank you because that person is clearly very influential. When you look closer at this tweet and really start to think about it with an unbiased point of view you start to realize the incredible act of deception here. And I'm not just talking about deceiving the president to reply, I'm talking about the 100,000 people that signed up to follow this account thinking that it was a group of like-minded people located in Tennessee, when it was actually a malicious puppeteer sitting somewhere across the globe. Twitter is under attack. It's a beautiful platform made for free, open discussion about different ideas. But the people that are attacking Twitter every day are trying to control the discussion so they can ultimately control those ideas. The goal is to promote tribalism and sow discord in society. This is the second part of a three-part series we're doing on SmarterEveryDay about attacks against our social media platforms. And we're gonna address today what Twitter calls platform manipulation. Today we're gonna look at data and we're gonna talk to experts face to face so we can understand what these malicious people are trying to do, how they're trying to do it and the most important thing of all, how you can get smarter so that you don't take the bait?
We're going straight to the source so let's go to Twitter Headquarters in San Francisco and speak to the team in charge of protecting the platform against these types of attacks.
I'm Destin.
I'm Del.
Del, nice to meet you. So you're over Trust and Safety, is that right?
That's correct.
At Twitter and trust and safety in general is keeping people safe on the Internet.
That's the goal.
Okay, first question, are we under attack on the Internet? Are there people actively trying to do things to change the way we think?
Absolutely, right. That's what advertising is.
Yeah, good point, yeah.
And even when you get past advertising there's this whole fascinating world of things that people do and the reason that I kind of pause when people say attack is that people who are trying to get the word out about something they believe in or who are trying to raise awareness of something, engaged in a lot of the same behaviors that you might think of when you're actually talking about somebody who's trying to maliciously skew the information available on a given topic.
There's another issue you're probably hearing a lot about on Twitter these days and that's accounts getting censored or suspended. The first reason is Twitter's value system. You can go read about that on their website if you want but I'm just gonna tell you straight up, that's beyond the scope of what I wanna get into in this episode of SmarterEveryDay. If Twitter is censoring people because of their ideas, Twitter has to own that. However, on the flip side I also need to understand that there is a battle going on behind the scenes to try to keep the platform usable and free from manipulation.
If you were to make your own social media platform you'd have to make a decision between two extremes. On the one side you could run it like the Wild West, which kind of sounds nice at first until you realize that anything goes and stuff starts to clutter up like your spam folder. On the other side you could be a totalitarian but that's not good for liberty or free expression. So the trick here is to do a balancing act. You wanna make the most amount of people on the platform happy but at the same time have just enough rules so you keep the platform usable.
So back in the day we had the trending topics and the original way that we did trending topics was that it would be what are the things that are being talked about the most on Twitter? In a way, they were under attack, if you will, because the top trend every day was Justin Bieber.
Yeah.
We were under attack.
The Beliebers were...
Specifically the Brazilian Beliebers (man laughing) who wanted Justin Bieber to come to Brazil.
Really?
So what they would do is every morning they would wake up, they would log onto Twitter and they would start tweeting. @justinbieber I love you, come to Brazil one. @justinbieber I love you, come to Brazil three. And they were doing the number because we'd already made the fix that you couldn't do exactly duplicate text. So now they were adding a number onto it. And you would look at these accounts...
It sounds like a joke but it's not.
It's totally not. These were real people. This was not automated. We ended up going, this is not actually a reflection of what we want trending topics to be. So we made the change, the first change, which was we're going to actually rather than just saying, okay, well this is the topic that's being mentioned the most, we're going to shift to trying to really focus on how fresh a topic is?
Now, these fans weren't automating their tweets but automation is actually allowed by Twitter's terms of service. This leads to all kinds of really fun and useful automation accounts, Year and Progress, Medieval Death Bot. A guy named Jeff and I run an account called WhoIsInSpace and we occasionally operate that in autonomous mode. But these bots can get out of control quickly. If they're allowed, how do you filter out the bad ones?
So we've all seen these nasty bots, right? They've got suggested pictures on there and they try to get you to click a link so they can get all your money. I used to see these all the time but I don't really see them anymore. I'm freaking impressed that you guys are doing as well as you are because I guess it was two years ago I could see all this bot content and you've fixed that, at least in my feed. So I imagine there was a great amount...
Hopefully not just yours.
Yes, but there was a great amount of engineering effort that went into that, I'm assuming. Those bots had a certain countermeasure to get around your gates and safeguards. You fixed that so now the bots are doing counter, countermeasures. And so now what you're about to do is a counter, counter, countermeasure.
Eventually you just get up to an absurd number of counters. There are all sorts of additional things that you can do, speed bumps that you can put in to do little checks to test what behavior happen when you try X. So for example, there's an account and we're just like, something about this doesn't seem right. Maybe it's automated. We can throw a CAPTCHA at it. Did it pass the CAPTCHA? It passed the CAPTCHA, that doesn't mean it's not automated. It could mean that they've just purchased, having somebody who's going to do CAPTCHAs that are thrown at it. Okay, let's go ahead and have them confirm that they have access to the email address associated with the account. We're gonna send them a password reset link. Did they get there? How fast did they get there? How long did it take them to verify that, right? So you have to maintain access to email addresses associated with your account. You can't just use burner emails. You have to be able to go back and log into it. And look, friction can take any number of forms, right.
That's the term you use, friction?
Yeah, friction. And there are things that are necessary friction. There are things that are unnecessary friction, right. There's things where we really need to make this experience easier for people.
I personally experienced this friction that Del is talking about. I recently signed up for a new Twitter account. It's for something I wanna do in the future and I went to tweet for the first time after two weeks and the account was already suspended. I hadn't done anything so why am I suspended? That's kind of weird and irritating.
And attackers are trying to manipulate Twitter by using computers to automate the sign up process. Twitter's forced to use artificial intelligence to defend the platform because these computers never sleep and there's not enough humans to do the job. The problem is these bad guys are starting to make themselves as authentically human as possible in order to sneak on the platform. Heck, in some cases they are human. Add to this deception the sheer number of signups and you can start to see how it's possible to miss some of these and let them slip past the defenses.
Speaking of defenses, after the first episode in this series I was contacted by the NATO Strategic Communications Center and yes, I'm talking about the North Atlantic Treaty Organization. I was given a report written by Sebastian Bay who works with the NATO Strategic Communication Center of Excellence to track all kinds of stuff related to NATOs interests.
Okay, I'm on Skype with Sebastian Bay. Happy 70th birthday of NATO, by the way. Sebastian has worked on this report, Manipulating Social Media, The Black Market for Social Media Manipulation. Do you mind introducing yourself so everyone knows?
Yes, my name is Sebastian Bay. I'm a Senior Expert with NATO Strategic Communications Center of Excellence in Riga, Latvia.
You've basically infiltrated these black markets and you understand how they work at this point, right?
So we've mapped the market. We've looked at all the different resellers out there. We've looked at prices. We've looked at methods. We've basically seen what are they selling and how are they trying to manipulate social media? How easy is it to buy a click or a like out there and how effective are they?
When it comes to Twitter profiles what are you seeing on the black market?
So I should take one step back, so we've been looking at Twitter for quite some time actually and we published a report called Robotrolling, which is a quarterly product where we look at manipulation of Twitter when it comes specifically to NATO issues in the Baltic's. And when we started looking specifically at this topic two years ago, as much as 70% of all activity relating to NATO and the Baltic's was actually manipulated. It was done by inauthentic bot accounts. That has gone down significantly though. Today it's down to the low tens.
What? So 70% of all, on that specific topic online, because yeah, why are people tweeting about NATO? And the answer is.
And the Baltic's, yeah. So when we looked at NATO and the Baltic's and especially in Russian, we could see that that environment was extremely polluted. Most of those accounts are very simple bots. They don't have a history and they're only there to inflate the account or to make a certain topic trend, etc. Twitter's done a lot actually to sure-up this but we can still see that these very simple accounts and these very simple bots can live as long as five to six thousand tweets before they get taken down.
So in your estimation, the new countermeasures that Twitter is developing, is it being effective?
Yes, in terms of it has removed a lot of the simple bot activity. But we're seeing different kinds of manipulation instead where we see more of these anonymous accounts that are pushing content. So these are more developed inauthentic accounts that are used for different kinds of social media manipulation. But when it comes to Twitter's response to these simple Twitter manipulation, that has been effective. That has gone down over the past few years.
How do these fake people get around those countermeasures that Twitter's developing?
Loads of different ways but when we talked to these people, because we've interviewed people that sell these things, they'd say that when social media platforms put new regulation into account, into effect, it takes them about two or three weeks to counter those new speed bumps that they put in, if you wanna use that word. And they do that by being inventive, simply by trial and error. If they put in a CAPTCHA service, well, they do ways to break that. If they put geofencing on IP addresses they use proxies, etc., etc.. So it's become more difficult but we're seeing, so far we don't see any limitation on the supply side. So you can buy fake activity on Twitter for an example within 15 minutes if you have a PayPal account. Many of the people we have spoken to that do these engineering, they are people that live in developing countries but also in developed countries.
So the people that buy and sell these services on the black market are obviously intelligent and they're there with intent. There's automated bots, there are actual humans. What is it they want to do on the platform? To try to figure this out let's go back to talk to Renée DiResta. She's an expert in understanding social media manipulation. So my understanding is the game on Twitter is to try to amplify inauthentic behavior using a bot or an account that you've taken over or something like that, is that how the game works?
Yes, so what you wanna do is you wanna get your message to reach the greatest number of people. And so you'll have bot accounts that just do it in an automated way, but Twitter's gotten better at detecting those. So then you have people who literally sit there and operate the account some percentage of the time. Sometimes it's an agency, sometimes it's just a regular person. One of the main things that Twitter's great for though is there's tons of journalists on it, which means if you wanna manipulate a story out of the social ecosystem and get it onto the nightly news, Twitter's a great way to do that if you wanna get it into a newspaper. 'Cause once something gets a sufficient amount of attention on Twitter,
Critical mass.
yep, then the press will pick it up.
Clearly this is news, I should put it on my news channel that I have.
You got it. And it's hard because for a long time, especially through 2016 you would see stuff would just trend, just show up in the trending algorithm. And then if the press wrote about it it was giving it more attention. If they were debunking it they were still giving it attention. And then if they didn't write about it you'd have this next wave of accounts speculating about why the mainstream media was ignoring it? So there was really no winning. The whole game was if you could get something to critical mass on Twitter you could just gain that attention economy and get massive lift, reach millions and millions of people.
Trying to get something to trend in the news is more than just trying to make money on the platform. This is a malicious, clever thing. Sebastian said he is seeing those types of services reflected in the black markets, something he calls meta manipulation. I understand everybody now thinks about the Russian bots and the 2016 election but I have to understand that the countermeasures have been developed and people know about that now, so they're already moving on to a different type of manipulation. What's going on now?
And what we're seeing now is a lot of focus on meta manipulation. So that means that you inflate the view counts to trick the algorithms of different social media platforms to make your own content trend on those platforms. So we've run experiments where we've bought bot activity. When we do that, that attracts real viewers because the platforms think that they are trending. So this sort of meta manipulation, that's what we're seeing now. That's what many of these resellers are promoting.
So this is alarming obviously but there's something that's even more disturbing about this. These accounts are the low tier accounts. There's a higher tier account that exists. That one is more expensive to create, more expensive to operate and is more devious. It's specifically tailored to misrepresent to misrepresent a specific group of people.
Accounts like this may actually have several thousand actual human followers.
And you can actually buy almost any kind of custom made account. If you want a Swedish female who's been online for 10 years that has several friends with specific interests you can request that from these vendors that sell accounts.
Really, so you can tailor the specific demographic of the inauthentic account that you want to use for manipulation?
Sure.
If you're trying to attack the African American community and create divisiveness in the African American community you can buy that kind of account online, or if you're trying to attack a specific political group, you can do the same?
Yeah, it's exactly like that.
These specialized accounts are hyper interesting social media weapons. Think about it, they're like a Trojan Horse because they can get you access to places that you're not expecting. They're like a false flag operation because they're misrepresenting what they are in order to blame something on somebody else. And with these weapons you can perform precision strikes because you can target a very specific individual on a very specific issue.
Twitter is also an incredible way to reach influencers because celebrities are all on there with millions of followers, and if they retweet you your message just hits millions of people or to reach the press, which is also on there. And so that's how you have, if you get something trending on Twitter reporters will pick it up and then it winds up on the front page of The Times the next day.
So the idea is to trick influential people and then use those as a vehicle to get this inauthentic information out there?
Yep, Twitter is a...
So if you have more followers you have a higher standard, you have to be even more serious about what you tweet.
Absolutely, yeah. You've gotta be really careful. When you think about if you retweet something that's wrong you have just pushed it out to potentially hundreds of thousands of people and they're not gonna see you necessarily push out the correction.
So when Twitter finds some of the bad guys they collect all the data and then they'll do one big data release so you can pour through it and see what's going on? People in academia are studying this stuff to try to develop AI methods to detect these people and counteract them. It's a cat and mouse game.
One of the most interesting thing that's happened is congress released a report with thousands of the actual Twitter account names on them. If you pour through this data you'll figure out that there's three basic categories. There's accounts on the left, accounts on the right and then there's this weird local news category. I think it would be helpful to explain what an attack against you would look like? It's gonna come in the form of a retweet from a friend. It's gonna be something funny that you resonate with. There's a single moment you need to be looking for and this is the tell, because you're not going to detect these accounts. Even the CEO of Twitter retweeted one of these accounts when they're tweeting all this stuff, the feel good stuff. But the moment you're looking for is a nudge, the nudge to think a negative thought against another group of people. If you belong to a certain demographic set or like a value system or belief system and you see someone saying something like, oh well, they always think that, we don't do that, or something like that, that's it. That nudge to think something against them, that's what you're looking for.
With terrorism there's a more clearly defined set of things that are bad or there's a clearly defined organization.
'Cause there's a literal battlefield.
There is, yeah. (lady laughs)
And this is the battlefield of the mind.
Right.
This is a battlefield of the mind where the enemy hides behind civilians. And the enemy has the added advantage that when they are revealed as being an enemy you basically start to distrust those around you, which was one of their goals to begin with. They're trying to get you to not trust people in your own community. It's very subtle but it's very complex.
Right now there's a really strong fear of false positives leading to unfair take downs. I mean, you're a content creator, right? This is never a thing that you wanna see happen and so the question becomes how do you send up a flair saying we think that this account might be something that we should have a look at? We think this account might be inauthentic. We think this account might be foreign, pretending to be - because the minute -
something else.
you have a false positive,
The Internet explodes.
The Internet explodes because Chubbywabs385 or whatever, surfing engine 385 gets upset. If you rewind to what I said at the beginning of the video and think about the suspension, this yes, some of these are as a result of management-making policy decisions about how people are interacting on the platform and things like that. But there's this other category, engineers are literally taking out malicious accounts at a rate of over 12 per second. Against an enemy that's constantly adapting, I can tell you, these tools they're developing are not perfect so it is not unreasonable to expect that every once and a while, unfortunately, there's gonna be an innocent bystander caught if you watch this video and your whole takeaway is, well, I'm just not gonna trust anybody on the Internet, then there's an argument to be made that I just played into the enemy's hand by making this video. If they hide behind civilians and then when we figure it out the result is to not trust anybody around you, that is not good.
These are our neighbors. That is next level psychological warfare that they're hitting us with. I'm gonna podcast and my cohost Matt and I have this phrase that we use a lot, political grace. If the point of these attacks is to divide us then proactive, intentional unity has the potential to be an effective countermeasure on the personal level. Maybe we should just not assume the worst of each other. And that extends to the people that are developing these countermeasures at Twitter. Their job is to perfectly protect a platform and never make a single mistake. And frankly, it's an impossible job that never ends.
Do you feel like in some cases you're overmatched?
- I think that we always try to assume that we are not as advanced as the people who are trying to attack us, because assuming otherwise basically puts you in a position where it's easy to develop blind spots. If we are always assuming that this is an arms race and that we should never be patting ourselves on the back going, we got all the countermeasures in place, we're always going to assume that it's evolving, that it's evolving quickly and that we have to stay on top of it.
And that's really the only way I noticed on the last video you were commenting on the video, specifically to boost it on the algorithm, that means you get it. Thank you so much. Also, I'm gonna do something weird. I've never done this on SmarterEveryDay but I'm gonna tell you about my Twitter account. I normally don't talk about my Twitter account but come on, I mean, if there's ever a video to tell you about my Twitter account this is it. So I think if I align this neon head up right there you should be able to click on it and go to a page where I've set up all my Twitter accounts. I've got @smartereveryday, which is the channel, of course. We got @destinsandlin and a couple of other surprise ones I don't normally talk about. So there you go, you can go click on the Twitter stuff. I hope this video earned your subscription. This is a lot of work, that's an understatement so I hope you enjoyed this content. Please feel free to share it. I'm Destin, you're getting smarter every day.
- Have a good one, bye.