What Social Networks do with your data (Part 2) [TL;DR warning]

I've been a fan of Project Veritas for quite some time, but now they have gone to the next level and exposed how Twitter abuses the data people entrust to it. In this blog I will break down what Twitter and most certainly other big data "Krakens" like Facebook, Google (and its parent company Alphabet), Microsoft, Apple and many more do with the data users grant them. This episode focuses on the first of the undercover videos that Project Veritas has filmed of Twitter employees telling in what they would never want their users to hear.

Note that the actual video starts at about 01:42. You can find it here

Picture sources: here and here

Embedded (Please go to 01:42):

The truth of Twitter betrayal: Part 2 [TL;DR warning]

The lie of "we have no bias"

The video begins with Mo Norai, former Content Review Agent of Twitter claiming that:

Twitter was probably about 90% anti-trump, maybe 99% anti-Trump.

At this point, it is unknown whether he is talking about the Twitter staff, or about the users, though I would say that if half the country elected a president, the amount of people hating him could not be above 90%, assuming the rest of the world isn't 100% anti-Trump either which I know for a fact from my surroundings. It would also be congruent with what Clay Haynes said in the last part of this series (here), to quote him again:

In fact, we've had internal reviews about that and we... I wasn't the only one that basically said that if we let that maniac... Something along the lines of, if we let this maniac continue, we would have a hard time finding a new job.

Note that Haynes and Norai are in completely different departments: One reviews user content, the other manages networking. As they say the same thing on different occasions and do not share the environment, it is likely true. But it still widens: Mihai Florea, Software Engineer of Twitter says:

It's really hard to decide what to do about Donald Trump. Half the people want to ban him, half of the people want to keep him.

Again, it is not clear yet if this is directed at Twitter employees or Twitter users, nor does every anti-Trumper want him banned. It could be either, but since the previous quotes were most likely talking about a huge majority anti-Trump, either he works at a place where the pro-anti quota is balanced or it is attributed to the users. It may also be that others have been overstating their parts: In case of Haynes, it is possible that a lot of people support Trump but don't want to openly say so fearing they will be fired, and in the case of Norai it does seem like a loose number he simply estimated. Until further light is shed onto this, we must not judge.

The Video jumps for a moment to Clay Haynes, when asked if everyone at Twitter felt like [here the PV journalist was interrupted], the answered:

Pretty much.

Note that in this case we don't have explicit context yet, but Haynes continues:

Yeah. I mean, you got to go to Google in order to find the Conservative.

At the very least we now know that he was talking about Progressives. In the last part, quoting him again:

I'm a bleeding-heart liberal. I think it comes with the territory.

Which would make absolute sense. in this context, plus the clips are in the same place and shot from the same angle. I would assume that he means that at the very least the others are "bleeding-heart liberals" as well, and in the worst case anti-Trumpers.

The video continues with Olinda Hassan, Policy Manager of Twitter's Trust & Safety department. She states:

I'm in Trust & Safety, I do all the policy work, safety, policy. I do... I'm in a controversial team.

So far nothing too shady. She elaborates:

I'm in the team that everyone says a lot about, yeah.

Project Veritas gives some context to this: Recently Twitter changed their Rules in order to, according to the post which you can find here, "reduce hateful conduct and abusive behavior". It does talk about "close coordination with experts on our Trust and Safety Council" in the opening paragraph. Now we know who we are dealing with. On with the video: When asked how Twitter keeps "certain things" off one's timeline, she answers:

We're trying to down rank it, but you also need to have control of your timeline.

Now what could they be talking about, considering it is most likely that Twitter's upper ranks are filled with "bleeding-heart liberals"? Maybe conservatives? Pro-Lifers? Trump-fans? Capitalists and anti-[crony]Communists? Anyone to the right of Mao and Stalin? At least she gives the PV journalist good advise in controlling her own timeline. But from previous quotes from Haynes we know that Twitter in fact has the power to down rank and even delete or create whatever they want. To quote Haynes again:

We have full access to every single person's account, every single direct message, deleted direct messages, deleted tweets.

This will be more important later on, so remember it well. The PV journalist continues and asks why "people like Cernovich [who is not in the favor of the left] and stuff" even though they are muted still show up all the time. Hassan answers:

Yeah, that's something we're working on. Yeah, it's something we're working on, where we're trying to get the sh*tty people not to show up. It's a product thing we're working on.

We know of one more person that definitely does not like people like Cernovich, and we also know that Twitter is working on methods to stop them from speaking out. Anyone else hear that faint sound of the 1st Amendment being murdered in their ears? But there is still more, she talks about it being a "product". Who will it be sold to? The users, the Twitter administration, companies? Or maybe investors and governments? Will it be a feature in the future for whoever has the money to buy the restriction of free speech of certain people? I can certainly see a market for that.

After this, the video goes back to Mo Norai, elaborating on what kind of people may be described by Hassan's "sh*tty people":

Let's say if it was a pro-Trump thing and I'm anti-Trump, I was like, I banned this whole account. It goes to you, and then it's at your discretion. And if you're anti-Trump you're like, oh you know what, Mo was right, f**k it, let it go.

It is too early to make any judgment on that, as this is a hypothetical scenario, but it looks like in this case Norai talks about himself as if he was just a random user, reporting a random pro-Trump account, with the Twitter employee of the department he was in simply following his bias and banning the account. At the start of all of this, he stated that Twitter was 90 - 99% anti-Trump. So we can imagine what would happen if these people no longer follow the actual rules, but their own bias. Combine this with Hassan and her new rules to get rid of "sh*tty people" and we have what is know as selective enforcement of rules: Anti-Trump people may not be banned even though they didn't follow the rules, while pro-Trump people may be banned even though they haven't done anything wrong. It's like saying "everyone of you is guilty and has to be banned, except for the ones I like". Norai continues:

On stuff like that it was more discretion on your view point, I guess how you felt about a particular matter---

This further supports the suspicion of the actual content review agents not following Twitter's own rules. The following conversation with the PV journalist continues like this:

PV journalist:

Oh, so it wasn't automated. It was---

Norai:

No, no---

PV journalist:

---a user end services person would deem it pro-Trump and take it down?

Norai:

Yeah, if they said this is pro-Trump, I don't want it because it offends me, this, that. And I say I banned this whole thing and it goes over here and you're like, Oh you know what? I don't like it too. You know what? Mo's right, let it go.

Notice the similar hypothetical scenario, especially the end of it: "You know what? Mo's right, let it go.". This seems to be a common phrase that he says whenever he talks about this subject. I might be wrong, but I also find myself using certain similar words, phrases and examples whenever I talk about a subject. Norai continues:

That's it. You're like, Mo was right, let's carry on. What's next? Keep it coming.

This sounds like there are certain people that are basically the "account-traffickers" for anti-Trump content review agents, seeking out pro-Trump accounts and reporting them, and it sounds as if they are in direct communication with each other. The following part only confirms the bias of Twitter's content review agents. When asked whether something flagged goes to him [or most likely his department] he again states:

Correct, and then, you know, you're looking at it and you're like, oh hey, this is pro-Trump... I don't like it.

But then it gets to what I assumed earlier: When asked if a lot of "left leaning or liberal stuff" would be dismissed unchecked, Norai answers:

It would come through checked and then I would be like, oh you know what? This is okay. Let it go.

So now we have the statement of selective rule enforcement. But wait, there's more: Asked about "unwritten rules" about what Twitter allows and what not, Norai answers:

Very. A lot of unwritten rules, and being that we're in San Francisco, we're in California, very liberal, a very blue state. You had to be... I mean as a company you can't really say it because it would make you look bad, but behind closed doors are lots of rules. Like, hey, you gotta do this this way. Or something like that. It was never written, it was more said.

The unfinished "you had to be" in there sounds like Twitter has been looking for people with exactly that bias to review user content, most likely the location of a department like that also factors into it. Why? Because they knew it what would happen and it fit Twitter's agenda. This would then explain the blatant anti-Trump sentiment that others like Haynes have been talking about: It's by design.

Next the video goes to Pranay Singh, Direct Messaging Engineer of Twitter. He is asked why Twitter won't verify Julian Assange:

I should ask about this, that's a really good point.

and when asked why they took the account down, he answered:

I didn't know we did that.

and finally:

Oh, it might be the U.S. government pressuring us.

It is clear that in this specific case, he does not know, but what he says next is important. He states:

Yeah, they do that.

and confirms again upon being asked. This tells us that the U.S. government has in the past had influence on Twitter, most likely still retains some. Given that Trump is so much in the firing line of Twitter, it seems that this does not extend to him, although he may prevent his account from being banned this way, it is unlikely that any pressure is coming from Trump or any pro-Trumper in the White House. This leaves the neutral, or given the focus of Twitter most likely the Anti-Trump people in office, most likely people that are now out of office as this could have easily gone on in the past under Obama as well. But I have another idea on who he may refer to, coming from the previous part and quotes from Haynes:

We're more than happy to help the DoJ in their little investigation.

Is it possible that Twitter is being pressured by the Mueller witch hunt team? When asked further Singh states:

Because they don't like people messing with their politics, and he has shit on a lot of people.

In essence, parts of Twitter's administration are being blackmailed by anti-Trumpers. But who is he referring to? He does seem to have a specific person in mind. This is just my wild guess, but taking all of this into consideration it looks like a high-level anti-Trumper, maybe Mueller himself, is blackmailing all private messages, deleted tweets and more out of Twitter in order to get Trump out of office. For more on this, again, refer to the first part of this series here. In the conversation, he confirms this again. When asked if he gets requests from the government to take down celebrities, he answers:

Oh, all the f**king time.

And that's it. Remember how I talked about the "product" of Twitter restricting the free speech of people for money [and blackmail it would seem] mentioned by Hassan? Looks like I was right on that one.

Next the video goes to Abhinav Vadrevu, former Software Engineer of Twitter, being in the same department as Mihai Florea from earlier. He states:

One strategy is to shadow ban so that you have ultimate control. The idea of a shadow ban is that you ban someone but they don't know they've been banned, because they keep posting, but no one sees their content. So they just think that no one is engaging with their content, when in reality, no one is seeing it.

Is this said "product"? I may be wrong, but it surely looks like the puzzle peaces fit here. He continues to describe the method of "shadow banning":

You just sort of turn off all the features for them. So like, they still see everything, it's all there. You can like it, you can favorite it, or you can, like, retweet or whatever. But at the end of the day, no one else interacts... No one else sees what you're doing. So all that data is just thrown away. It's risky though.

When asked why he continues:

Because people will figure that sh*t out and be like... You know, it's a lot of bad press if, like, people figure out that you're shadow banning them. It's like, unethical in some way. You know? So, I don't know.

This is congruent with, again, Haynes from the previous part, who stated:

[...] I don't like being a part of the machine that is contributing to America's downfall.

It seems as if Vadrevu shares this. These people know what they are doing is wrong, and even though they may not like it, they still move forward. Out of need for money, or maybe just to get back at Trump and his movement. He continues:

In the past people have been really, really pissed off about that. And even people who haven't been shadow banned have called it, like, a really terrible thing to do. So, yeah, it's a risky strategy.

When further asked about how "shadow banning" works he answers:

Yeah, yeah. I definitely know Reddit does this, but I don't know if Twitter does this anymore.

I guess Project Veritas has their next target: Reddit.

The video moves on to Conrado Miranda, a former Twitter Engineer. He is asked about Twitter banning people like Trump supporters or Conservatives and confirms that this is not just a rumor. When asked how it is done, he answers (not after a moment of seemingly thinking about whether he should talk about it or not as he stops moving food towards his mouth for a brief period):

Okay. Basically, Twitter is, like, a huge set of, like, the tweets and a sh*t load of filters on top of it.

and then he states about the banning policy:

That's one of the filters on top of it.

So not only is Twitter actively going after Trump supporters or Conservatives, that is just one tiny part of the manipulation. He continues:

That's about it. Like, that's as much as I know about...

So the million dollar question is: What is going on that we don't even know about yet. Miranda elaborates:

Yeah, like... Imagine, like, you have all the tweets ever that happened, like, from the beginning of time in reverse chronological order.

and:

So like, until we get to the actual system that, like, ranks the timeline, we just, like, we have a bunch of filters removing some tweets. So like, we have a stream of tweets and we have, like, one filter just kicking out some of them.

In other words: An ominous algorithm is looking for certain signals in tweets and just deletes them from other people's view. This is most likely a part of the shadow ban, where the actual implementation of it is a filter with very, very harsh rules to it. To the point where it filters out everything and nothing shows up from that account for the average user. The video goes back to Pranay Singh, stating that:

Just go to a random (Trump) tweet, and just look at the followers. They'll all be like, guns, got 'Merica, like, and with the American flag and, like, the cross. Something---

and continuing:

Like, who says that? Who talks like that? It's for sure a bot.

After a small conversation in between, he is asked if he can get rid of "rednecks" and he answers:

I wouldn't know.

but then corrects:

Umm, yeah.

And when asked how he continues:

You just delete them, but, like, the problem is there are hundreds of thousands of them, so you got to, like, write algorithms that do it for you.

This is what the Software Engineers do. And we also know their bias. When asked about how to differ between people and bots he answers:

You use machine learning.

One of the PV journalists makes the point that the algorithm just seeks out people with, in this case, American guns, and simply deletes them. Singh responds:

Uhh, yeah. It's actually how we do it, yeah. [...]

The PV journalists continue to make examples of algorithms that search for keywords in the names of people or find their political opinion, to which Singh answers:

Are you sure you're not a programmer? That's exactly how you do it.

He elaborates:

You look for Trump, or America, or any of, like, five thousand, like, keywords to describe a redneck.

Note that while initially the question was on how to differentiate bots and people, the conversation no longer focuses on that but only on people. He continues:

And then you look and you, like, parse all the messages, all like the pictures, and then you look for, like, stuff that matches, like, that stuff.

Or in other words: Twitter uses machine learning and AI technology in order to find out a person's character from text and pictures (which we know from Haynes out of the previous part that Twitter keeps all of, even deleted ones and pictures you uploaded, but then canceled or replaced). Singh continues:

And like if it, so you, like, you assign a value to each thing, so like Trump would be, like, 0.5, a picture of a gun would be like 1.5, and, like, if it comes up the total comes up above, like, a certain value, then it's a bot.

It seems as if the real agenda here is not looking for bots. It is to classify people who post things of, in this case, Trump pictures and guns as bots and use this as a pretext to ban them. He is asked against who the majority of algorithms is, and it should be no surprise that he answers:

I would say majority of it are for Republicans, because they're all from Russia and they wanted Trump to win, so...

and when asked about it targeting conservatives he confirms. It is, the same as with Haynes, not my skill to judge people from just a few short clips, but it looks as if Singh is a mislead SJW that truly believes the evil Russians are behind everything. It seems that he legitimately cannot believe that there are people who are pro-Trump, pro-Gun. He is trapped in the echo-chamber that is the liberal hive (I made a blog on that a while back, you can find it here), and he most certainly, like Haynes, would want to help Mueller and his team impeach Trump.

The video continues with Steven Pierre, another Software Engineer from Twitter who is asked about a "secret project" that would "help people communicate better". He answers:

That's the goal. So what I'm working on. Okay, this is the goal of what I'm working on, ultimately. The goal of what I'm working on is going to make it so everything on Twitter is dynamic.

After a brief info on machine learning from Project Veritas, Pierre continues:

Every single conversation is going to be rated by a machine and the machine is going to say whether or not it's a positive thing or a negative thing. And whether it's positive or negative doesn't look for the content, it's more like if somebody's being aggressive or not. Right? Somebody's just cursing at somebody, whatever, whatever. They may have a point, but it will just, like, vanish.

Note that in the machine learning employed for this kind of operation, the machine does NOT decide positive or negative. These are preset from human input. It only detects and rates, but doesn't judge. The definition of "positive" and "negative" is left in this case to extremely leftist anti-Trump Twitter staff. However, what the machine does in response is, as it sounds, banning these people, and from what is possible it could also issue shadow bans. Pierre, when asked if this is applied to people with certain mindsets, he answers:

It's not going to ban the mindset, it's going to ban, like, a way of talking.

What are you willing to be that this is exactly the product Hassan has been talking about? What would it be? An algorithm using machine learning to interpret text and pictures that [shadow] bans anyone who doesn't fit with the leftist, progressive agenda of its programmers and supervisors. It is my guess, and as an amateur programmer myself call it an educated guess, that this is what Twitter might in the future, or perhaps just now is presenting to governments, especially in China and Islamic nations, and the controller-elites.

You like Trump? BANNED! You want your 1st and 2nd amendment? BANNED! You fear crony communism? BANNED! You want to simply talk about reality? BANNED. You don't like unlimited immigration? BANNED! Anyone in the government or elite pays some money or has something on our staff and wants you gone in return? BANNED! You don't support Jihad? Islamists will break open your door and kill you, rape your wife and children, and then kill them. You are critical to the government? A few shady people punch you to a pulp in the dark - or just kill you when you open the door. Depending on whether or not shadow banning is still active, it might also be applied. This is what the future of Twitter looks like. Glad I never was there. Though you really don't need an algorithm to figure out my political agenda. It's written on the back of my channel and if you look closely in high resolution, even in the small avatar.

I'm sorry this was so long, but I want to do a blog per Project Veritas video, and there are just so many cross-connections in there. I hope you enjoyed this and thanks to anyone who was enduring enough to read this.