Pod Insights Decoder Recode Decode: Twitter's Vijaya Gadde and Kayvon Beykpour (Live at Code 2019) About this episode Vijaya Gadde, who leads the legal and trust and safety teams at Twitter, and Periscope co-founder Kayvon Beykpour, who's now the company's head of product, talk with Recode's Kara Swisher and Peter Kafka at the 2019 Code Conference in Scottsdale, Ariz.
In this episode: Twitter’s meeting with President Trump; CEO Jack Dorsey's level of contact with the policy team; cleaning the Twitter “cesspool”; could it operate without letting everyone speak?; its new policies around elections and anti-vaxxers; how its responses to abuse compare to Facebook’s and Google’s; does de-platforming people like Alex Jones reduce their influence?; does Twitter radicalize people?; how Twitter is trying to get rid of white supremacists; and false equivalency in content moderation.
Learn more about your ad choices. Visit podcastchoices.com/adchoices
Transcript 9,431 words · 12 speakers
0:01 Speaker 00 What if managing company spend took less managing? Introducing Amex Corporate, one intelligent platform connecting cards, expenses, and payments so you can keep work moving. Built to simplify, designed to scale. This is Amex Corporate. Get started at go.amex/ammexcorporate. Terms apply.
0:29 Speaker 01 You could try to find a perfect coffee to water ratio. Or you could press a button. With Keurig, each K-cup pod is precision-measured for a perfectly delicious cup. No effort, no grind, no math required.
0:46 Speaker 01 Great coffee without the grind. Keurig.
0:52 Speaker 02 Hi, I'm Kara Swisher, editor-at-large of Recode. You may know me as someone who will only be happy if Twitter cleans up its platform or gives me an edit button, but I need at least one of those things. But in my spare time, I talk tech, and you're listening to Recode Decode from the Vox Media Podcast Network. Today, we're going to play an interview from this year's Code Conference. Peter Kafka and I talked to Vijay Gadi and Kayvon Bakpour from Twitter. Gadi is the company's legal boss and also heads its trust and safety team, while Bakpour is the head of product. Gadi says solving Twitter's problems isn't as simple as kicking off all the bad actors as some of the platform's critics like to claim.
1:29 Speaker 03 I don't think anyone can convince me that bad things don't happen on private platforms. Bad things probably happen more often on private than public platforms. There is an advantage to being open, which is that everyone can see it and respond to it and understand what's happening. Now, there's a disadvantage to that as well, obviously, because you give people a platform. So you have to find the right balance of being able to provide a space for ideas to happen, but counter narratives to flourish.
1:56 Speaker 02 You can find full coverage of this interview and everything else from the Code Conference at fox.com/recode. But now, let's go to the Phoenician Resort in Scottsdale, Arizona, to hear Peter and me talking to Twitter's Vijay Gadi and Kayvon Bakpour.
2:10 Speaker 02 Thanks, everybody. You know, I was just looking at Twitter and there's a new drinking game with if someone says, "I'm gonna try better," and there's room for improvement. You'd be super drunk. You're gonna take a shot, yeah. Yeah. But you don't drink. I don't drink, but that's still, it's, all right. In any case, it's kind of funny. Anyway, I told, I warned Vijaya that she can't say any of those things onstage, but we're very excited. One of the, obviously, the other companies that's very important in the social media space, we wanted to have them all here, is Twitter. And so, we wanted to bring in two people who are critical to everybody focusing on Jack and his beard and eating habits. These people do the stuff. These people actually do the stuff. Kayvon Bakpor and Vijaya Gotti.
2:52 Speaker 02 Come on out. Start drinking. You know what you're not supposed to say. I mean, I'm glad there's room for improvement, but you know what I'm saying, you get my point. So, I think I want to start off, I really, really want to understand, you were in that meeting with President Trump, is that correct? Vijaya?
3:10 Speaker 03 I was in a meeting recently with Jack Dorsey, our CEO and President.
3:14 Speaker 02 And President Trump, but President Trump was also in the room?
3:16 Speaker 03 Yeah.
3:17 Speaker 02 Okay. Because it was the Oval Office. So, can you, what happened in that room, exactly? And in detail?
3:25 Speaker 03 In very specific detail.
3:26 Speaker 02 And how did it come about? How did it come about? And then, what happened exactly?
3:30 Speaker 03 It came about because, you know, a lot of tech company CEOs had met the President on several different occasions. That hadn't worked out for Jack to do that. And so, we had received a request from the President's team that would be nice if we could arrange a time for them to meet and talk about different issues. There was no actual agenda for them.
3:49 Speaker 02 They just, they called you and wanted to talk. I mean, he is your best customer, but, but when you, when you're, I'm sorry, that was too easy.
3:56 Speaker 03 It's okay.
3:57 Speaker 02 What did you, what did he want to talk about? And what happened in the meeting, precisely?
4:00 Speaker 03 We talked about a number of different issues. We talked about the platform, the use of the platform around the world. Jack also specifically talked to him about improving the civility of public conversation and how important that was.
4:15 Speaker 02 Yeah.
4:16 Speaker 03 So, it was a wide-ranging meeting.
4:17 Speaker 02 Sorry.
4:18 Speaker 03 And, you know, talked about a lot of things, but it was also about 30 minutes. So, how much can you really talk about?
4:24 Speaker 02 Yeah. When you talked about civility, did you, could you sit there straight-faced in terms of? She's doing it now. I know. She's doing excellent.
4:32 Speaker 03 I'm excellent at straight faces.
4:33 Speaker 02 It's true.
4:34 Speaker 03 I think it was really important for Jack to bring this topic up. Yeah. That was his purpose in attending this meeting was to impress upon the President how seriously we take these issues and also to talk about what we're trying to achieve, whether that's our work around election manipulation and preventing that from occurring or about improving the health of public conversations and why that's important to us.
4:56 Speaker 04 Can you talk a bit about the dynamic between Jack and you guys? Our perception is Jack goes out, he does listening tours, he talks to us. He does a lot of thinking about things and describing them as difficult. And then my understanding is then you guys have to do the work. Does he bring you an idea and say, "Let's fix civility. Let's have healthy conversations. You guys figure out how to do it." Does it bubble up the other way?
5:20 Speaker 05 I think that what you're touching on is that Jack's been instrumental in defining the topmost company priorities. So shortly after he came back to the company a few years ago, it was really his encouragement to get us focused on increasing the health of the public conversation as a topmost company priority.
5:40 Speaker 04 He brings that to you, says, "I want to do this."
5:42 Speaker 05 It's something that we discuss as a management team. As we flow down priorities across the company, Jack is most involved when he talks about the topmost company priorities. We then, as leaders, I myself on the consumer products side, we figure out how to translate that into a cohesive product strategy because for us, if we want to focus on increasing health or if we want to focus on making people comfortable talking in public, that sounds great as a top-level company strategy, but really translating that into a cohesive set of things that we can focus on is where we, as a cross-functional team, do all of our work.
6:14 Speaker 04 And then how granular does he get? I mean, there's a lot of, whether it's Trump or, like I was talking to Susan about YouTube and Crowder thing, oftentimes there's a discussion about, "Do you bring the CEO in to weigh in on whether this person should be demonetized or whatever the other punishment is?" How often is he involved in a discussion like that? Whether it's a very specific policy thing or this person needs to be off the platform?
6:36 Speaker 03 So I'll let Kayvon talk about product. But on the policy side, we have well-established policies and guidelines that we use. So I cannot remember a time where I've talked to Jack about a decision that we're making. I informed Jack about a decision we're making, so if he's going to read about it on Twitter or in the media, he's aware of it. What I do involve very deeply with conversations with Jack is where our policy is going, and how that ties very closely to the product and what we're trying to build. So those are the types of decisions that he would be very active in and that we engage as a team on.
7:08 Speaker 05 I think one of the amazing things about Jack as a leader is that he empowers other people, in particular his executive team, to be as autonomous as possible in driving their areas of the business. So likewise, you know, Jack doesn't get super involved in the day-to-day product decisions. You know, he empowers me and our peers in the sort of product engineering design research organizations to, you know, drive the strategy and make our decisions. And we get Jack involved when we can really use his input and help shape, you know, the work that we're doing.
7:37 Speaker 04 He's got two jobs too, right? So he can't... Does he have another job? That's what I heard.
7:41 Speaker 02 All right. So let's get to one of the things, the healthy conversations one. He talks about it, he goes like this, "We really want healthy conversations," the whole thing. Twitter, I think, arguably could be the place where a lot of unhealth... where it starts and it begins and ends with really unhealthy conversations. Now, I myself have called it many times a cesspool, although I love it at the same time. You know that. I use it quite heavily. I use it for all kinds of things I do. To me, the very nature of it is that it is a cesspool. It's never not going to be a cesspool because of the way it's built. Can you each talk about that from a policy perspective? How do you change that? And you don't have to agree with my cesspool assessment, but I think it's completely accurate. Is how do you do it from a policy perspective and then from a product perspective? Because what is built on is saying whatever you want, and when people can say whatever
8:26 Speaker 02 they want, they say terrible things.
8:28 Speaker 05 The way I think it's helpful to look at this is really starting with why we're even doing this in the first place.
8:33 Speaker 02 Yeah, I often wonder that, but go ahead.
8:36 Speaker 05 I'm sure you do.
8:36 Speaker 02 Yeah.
8:38 Speaker 05 Our purpose as a company is to serve the public conversation. And we think, starting with that and articulating why is important because it helps explain everything else that we do and some of the things we don't do. We think public conversation is important because ultimately it helps people learn, it helps people solve problems, it helps people realize that we're all in this together. I think some of the most important issues that affect civilization are increasingly going to be worldwide issues, not issues of a particular nation-state. Things like climate change and displacement of the workforce and all these sorts of things. I think having public conversation allows the world to better address these issues out in the open. And I think that starting point is sort of the thesis for why we believe Twitter needs to exist.
9:22 Speaker 05 Now, if that is your purpose as a company, you necessarily are creating a scaffolding that makes a ton of things really difficult, right? You're fundamentally predicated around letting people talk in the open. That's both terrifying and really powerful. And of course, then, from there, you have a bunch of existential crises. One, if that public conversation isn't healthy, no one will want to participate in the first place. And that tells you just how...
9:44 Speaker 02 Or they will.
9:44 Speaker 05 Well, they will, or it'll be miserable and they won't do it anymore, or they will and they'll feel awful doing it. All of these things, regardless of what the motivations are, are nevertheless bad because they get in the way of that purpose, living up to its potential. But that's one existential crisis that, again, we can go into detail on how does that actually impact our product prioritization and the work that we do, but that's one issue. And the other is, our service is predicated on people talking in public. Unlike many other forms of technology that allow you to stay informed about what's happening in the world, the fuel that helps people stay informed on Twitter is people talking. The atom of conversation is you tweeting, Peter tweeting, me tweeting, so on and so forth, all around the world.
10:25 Speaker 05 So if we don't make it easy for people, or comfortable for people to talk in public, then that fuel, those atoms of content that ultimately allow all of us to stay informed about the important things that matter to us in the world, goes away. And so, those two objectives are the basis of everything that we do. And all the product prioritization, all the policy work that we do, everything holistically ladders up into those two objectives in service of them.
10:50 Speaker 02 Did you talk about this because it then creates an enormous amount of problems for you because it creates, like you were talking about public, a public square? But you're not a public square. You're a private company owned by billionaires that's making a lot of money off of this. So are you a public? How do you look at it from a policy point of view? Because then it intersects with politicians who are like, if it's a public square, then we're going to regulate it this way or that way.
11:13 Speaker 03 I try not to think about it purely from a regulation point of view because I know a lot of people say, oh, you know, you're not subject to the First Amendment. The First Amendment doesn't apply. You're a corporation. And you're right, it doesn't. But what we're trying to do, as Kayvon mentioned, is serve this global public conversation. And how can you do that most effectively? For us, it's looking very closely at how we develop the policy framework, focusing on human rights, making sure we're not adversely impacting people's human rights. That's their physical safety. That's their right to free expression. That's their right to privacy. Thinking about that very closely. And then how do we allow as many voices to participate in that conversation? And so even our policy framework, though, over time, has had to change because it's
11:55 Speaker 03 clear that we didn't anticipate all the types of behaviors that we would see on our platform that we now see. And we've had to move very much from what we were, which was a platform that very much enabled as much free speech as possible, to one that is very cognizant of the impact that it's having in the world and our responsibility and our role in shaping that.
12:16 Speaker 04 Same question as asking Susan. Is there any way that you guys could operate not being a platform that allows anyone to speak, where there is some sort of hurdle that prevents, that you have to cross before you can start having a discussion?
12:31 Speaker 05 We could. I think that would be a fundamentally different product. That would be not as good for the world. I personally believe, and a lot of us at Twitter believe, that the, there needs to, like, global public conversation is fundamentally a good thing for the world.
12:47 Speaker 04 But most people who are tweeting, right? It's a very small minority of the users, right? Most people are passively sort of consuming it, right? You guys know better than me, but I'm assuming it's a fraction, a very tiny fraction. So you sort of have that already. Couldn't you say, look, before you start spewing whatever, whether it's sewage or anything else?
13:06 Speaker 04 We need some bona fides.
13:07 Speaker 05 I think that's, it's a convenient thought exercise to run, but if you think about the sorts of things that you and I as customers of the experience, when we open up Twitter to see what's happening in the world, when you think about the sorts of things that we're able to learn about, whether it's a breaking news event, or our favorite sports update, oftentimes you're hearing this information from people out in the world who are on the scene of the breaking news event, or have some perspective that is, has a variety to it that is not coming from...
13:41 Speaker 04 But increasingly, you guys are pushing people towards, "Look, get this NBA beat writer to comment," right? So the early stories of the Hudson landing, the plane landing in the Hudson, and someone snapped that photo, or the guy who was tweeting about the helicopters at the Obama raid, Osama bin Laden raid, right? Those are great anecdotes, and they're cool stories, and it's part of what makes Twitter cool, but it's not the majority of the...
14:05 Speaker 05 It is literally the majority. The minority of people talking on the platform are folks such as yourselves, who already have immense platforms of followers, who you can reach through a distribution platform like Vox, or Recode, or whatever it is. The majority of people who talk on Twitter don't have thousands of followers, they have tens of followers, or hundreds of followers, and it's that sort of flourishing ecosystem of people who can talk and find other people who have like-minded interests, to ultimately discuss the things that matter to them, that makes the service the service. And it's the reason why when the one-in-a-million situation, whether it's a plane landing in the Hudson, or the helicopters in a baba-ba, that then that piece of discussion happens to actually
14:47 Speaker 05 matter for a bunch of people outside of their circle.
14:51 Speaker 02 We're going to take a quick break now. We'll be back with this interview with Vij Jagade and Kayvon Beykpour after this.
14:57 Speaker 06 It's me, the
15:03 Speaker 06 We don't need the jingle twice. Stop it. This is about vacation inflation and how Priceline negotiates amazing deals on hotels, flights, and rental cars.
15:14 Speaker 06 Yeah, but I didn't mention that you can save up to 60% off hotels in the Priceline app.
15:20 Speaker 06 Fine. No one deals more deals than Please stop. Touche.
15:27 Speaker 06 Priceline.
15:28 Speaker 07 How about a latte?
15:29 Speaker 01 Maybe dark roast?
15:30 Speaker 07 Vanilla?
15:31 Speaker 01 Caramel?
15:31 Speaker 07 Blueberry?
15:32 Speaker 01 You could get them to agree on one flavor, or give everyone what they want. With Keurig, there's over 500 K-Cup pie varieties of fresh, flavorful coffee. No effort, no grind, no settling. Great coffee without the grind. Keurig.
15:55 Speaker 02 Let's talk about the challenges you each face with the product right now. Today's, we'll get to regulation because you're a smaller entity. I don't think you're facing quite the same challenges that Facebook and Google are with. Your company is not pulled into things, except when Ted Cruz gets mad about whatever he's mad about. But talk about the things that affect you. What is your major policy challenge right now for Twitter from your perspective?
16:20 Speaker 03 I think one of the toughest things confronting our industry is how to deal with the lack of trust in information, misinformation, whether it's deliberate or accidental, and how we address that. Obviously, people come to Twitter to find out what's happening in the world, and if they can't trust the information that's on there, they're going to stop using the product. So it's a very, very important issue. And I think the challenge with it is that it's not just about a policy. I can say you can't say anything untrue on Twitter, but I would never be able to enforce a policy like that. So how do we approach this issue in a way that we can use technology, that we can use the product itself, and create a scalable process for dealing with this type of content? So we're trying two different things.
17:04 Speaker 03 One, we launched in, back in April, which was a policy around misinformation specifically in the context of elections. So anything related to how you vote, how you register. And we rolled that out for the elections in India, which just happened, as well as the European elections, which just happened, and we were enforcing that policy, and we learned a lot from it. Including the fact that, you know, people make jokes about this stuff. And they didn't think it was going to be taken seriously. And all of a sudden, you know, a joke about when to vote got taken off the platform and they weren't very happy about it. So we're learning how to handle this. The second thing that we're working on is, you know, again, focus on this offline harm. Really, because those are the things that matter most to people. Is we launched in May, a new product intervention.
17:49 Speaker 03 When you search for things related to vaccines, you get directed to authoritative accounts such as the Department of Health and Human Services. So again, learning how we can best do this in a way that addresses this very monumental challenge in front of us.
18:02 Speaker 02 But not removing anti-vaxxers, just pushing them somewhere else. Your preference is not to decide.
18:09 Speaker 03 Our historical preference has been not to be the arbiters of truth. Many reasons for that, including a lot of our customers who don't want us to be making those decisions. And it's hard for us to do it at scale.
18:19 Speaker 04 But you are putting your thumb on the scale, right? You're saying, this is stuff we like, we're going to promote more of that and the stuff we don't like. And again, all the platforms have a version of this, right?
18:27 Speaker 03 It's not about what we like, but yes, we are changing the scales. We are trying to amplify content that we believe comes from credible sources, reputable sources. And we're going to, over time, be able to de-amplify the content that doesn't. But I don't think that that in and of itself is going to be enough. And we're going to need to be able to do more in this area. But I would say this is one of the biggest challenges.
18:45 Speaker 02 Why not just go there and say that's what you're doing? It's been such a, like, I remember when I saw Jack before the Alex Jones stuff that happened. I'd love you to talk to you that because we were in a thing and he's like, we feel like he should talk. I said, you're going to be taking him off. And he said, we feel like there should be voices that said, and yet you're going to be taking him off in, I don't know, three months. And he goes, no, and then he took him off. And I wrote him out and like, oh, you took him off. Like, it just was sort of, why not just go there to where you're going to end up anyway? What's, why drag us all through this mud?
19:15 Speaker 03 I think these are two very different situations. So I don't want to talk about that one in the context of this. I think this is an enormous challenge for the industry. And we're actually trying things. We're trying to figure out what works best. So it's not that we're not going to go there. I want to be very clear, we are going to do our best to address this issue. It's one of the most important things that Kiva and Jack, I talk about along with our CTO. And so this is where we're going. We're learning how to best do it. Every platform's not going to be the same. Every solution's not going to be the same. And we're also watching what our peers do and see how that works as well.
19:45 Speaker 04 So Susan was talking about the fact that YouTube can use all of Google's capacity to handle this problem. Facebook has, I think, Four times as many folks as working period, the workforce is four times as big as yours. Just technically and from a personnel problem, if these two giant companies who have the same problem are struggling to deal with it, how do you guys tackle it?
20:07 Speaker 05 Well, I think maybe connected to the previous conversation we were having, but also maps to this question, I think one observation that I would make as a critique of ourselves is that I think historically with this general problem space, we've over rotated on trying to solve the problem too much through policy and enforcement and not through also through product and technology. I think fundamentally, this whole space is a policy enforcement, product, technology, incentives problem. It's really does it a disservice to try and simplify it into just one of those buckets. And so I think first and foremost, we have had to have a shift internally around treating it as a holistic problem, that all of our peers across policy enforcement, product,
20:52 Speaker 05 etc. work on. And so I think that is why over the last year and a half, we've made so much progress on health and why I believe over the next year and a half, we will make even more progress because we actually are treating it holistically as a problem. When I think about what we've done on the information integrity side, you know, of which Vijay just mentioned, just election integrity, but when I think about what we've done on information integrity, on conversational health, we've brought to bear more product and technology solutions that start to tip the scales, to use the phrase that you did around deamplifying content that we think is likely to be against our policy.
21:28 Speaker 02 Give me an example of that, what's something that you did that?
21:31 Speaker 05 Yeah, I'll give you a few examples, maybe just starting from the beginning. One, we actually have simplified our policies to make them human readable, so that they're not just in legalese. We announced this last week, so that every one of our policies fits in a tweet. We actually enforce those policies more quickly, with more agents and more proactively. Right now, 40% of all of our enforcement happens proactively, and that's more than double what it was this time last year.
21:55 Speaker 04 So agents that's human beings, right?
21:57 Speaker 05 It's a combination of agents and machine learning, but in aggregate, 40% of our terms, 40% of tweets that we action for terms of service violations, we now action proactively. So that's one.
22:08 Speaker 02 Complaint, your complaint process. Meaning it's not through a complaint process? Exactly, not through a complaint process.
22:15 Speaker 05 Yeah. On the information integrity side, we've made a ton of progress around things like account compromises, fake signups, malicious coordinated activity. These are all sort of foundational aspects of sketchy behavior that happens on the platform that's unwanted. We've made dramatic progress there. You know, just as one example, we now challenge and block between eight and 10 million accounts per week that a year from now, a year ago, would get through our signup process that now don't as a result of the work. These are bots.
22:48 Speaker 04 These are people who are signing up, or robots that are signing up.
22:50 Speaker 05 It's a combination of all of that. As one other example, just one vector of abuse, which is brute password guessing, basically. We had, I think it was around 1.5 billion attempts per day before we started doing our work in this area, and we've now brought that down to 600,000 attempts per day. So these are all, the point you should take away from this is that this is such a comprehensive and complex problem, there is no single silver bullet. We have to address all the different holes in the foundation that we've had historically. And another example of this is on the conversational health side. A lot of what people consider abusive on the service doesn't actually violate our policies. Because what care finds abusive is different from what you find abusive and so on and so forth.
23:33 Speaker 05 And so, one of the things we've had to really step up from a product and technology standpoint is proactively deamplifying content that we don't think should be amplified. So whether that's in your conversation view, when you're trying to have a conversation with someone, or whether that's in your notifications filter, these are sort of amplification services where we, you know, people's speech ends up getting amplified as a result of our algorithms. Through the work that we've done over the last year, we've reduced abuse reports by 45% in the conversation view. We've reduced the number of blocks in the notification tab by 30% in the last year. And again, these are not, these are not like magic numbers, but they give us a sense of whether we are helping or hurting the things that were optimized.
24:15 Speaker 04 Do you have a sense of what happens to an Alex Jones, or a Milo, or someone who's on the platform you block? What becomes of their sort of social graph and power?
24:26 Speaker 04 When they're not on Twitter?
24:28 Speaker 03 I don't have a direct sense of that. Obviously, we talk to researchers all the time about the impact of different actions we take, different policies we have. And one of the things that has been a topic of conversation is what happens when you deplatform people. Are you increasing radicalization by forcing people to other corners of the Internet, whether it's private places, encrypted places, or other platforms that maybe have no policies around certain types of content? So it's a conversation we're having, but I don't have any specific statistics.
25:03 Speaker 02 Do you even worry about that? Let's keep this awful person here so they don't go over to parlay or wherever the gab or wherever.
25:08 Speaker 03 I worry about how to minimize radicalization as a whole and what role our platform plays in it. I don't worry about a specific person.
25:16 Speaker 02 Do you think about the radicalization? He asked Susan about that, and she gave essentially a non-answer whether she thought it was or not. Do you think Twitter radicalizes people?
25:27 Speaker 03 I think that there is content on Twitter and every platform that contributes to radicalization, no doubt. But I think we also have a lot of mechanisms and policies in place that we enforce very, very effectively that combat this. Over 90%, we've taken around 1.6 million accounts down for terrorism on the platform. Over 90% of that is detected by our own technologies proactively without any user reports. That's work that we've been doing for many, many years. Obviously, that focuses on all around the world. We have very specific problems in certain parts of the world that we are now addressing as a platform as well. But we have a lot of policies.
26:08 Speaker 03 We have a violent extremist group policy that has banned over 110 violent extremist groups. 90-plus percent of those are white supremacists or white nationalist groups. Including the American Nazi Party, the brow boys, the KKK. So there's a lot of work going on here that people aren't seeing. That's happening every day.
26:30 Speaker 04 A producty question for you, Kayvon. There was a period where Twitter was really trying to grow and they wanted to be as big as Facebook. And then there was a preceding regime said, "Well, there's not that many users, but there's concentric circles and people are using it." And now you guys are moving from reporting monthly active users to a different engagement metric. Do you have the sense that like pretty much everyone who's gonna be on Twitter is on Twitter and you're pretty much done growing and you have to sort of build a business based on the user base you have? Or can it keep growing rapidly? Or can it grow rapidly?
27:02 Speaker 05 I mean, the way I would answer that question is, you know, the purpose that I just articulated earlier, we fundamentally believe that that is a daily utility that billions of people could use. Billions of people don't use Twitter every day. But it is our aspiration that we can keep delivering value to customers that are trying to stay informed about what's happening in the world, inform other people about what's happening in the world and discuss the things that matter to them.
27:26 Speaker 04 So how are you gonna get someone who's not a regular user of Twitter to become one? It seems like everyone sort of looked at it and they've decided, "This is for me and Kara." Or no?
27:34 Speaker 05 Well, I'm happy to sort of walk you through some of the ways we're thinking about the product strategy. And first of all, the service is growing. So that, I appreciate the sentiment that, you know, we've sort of grown as much as we can and historically-
27:51 Speaker 03 I think you just said, "You're tiny and get used to it," but go ahead.
27:53 Speaker 05 Right.
27:53 Speaker 04 Flatlined, yeah.
27:55 Speaker 05 I think there's a few things that we need to do that fundamentally will deliver enough value that people will want to use their service more. One, you know, we talk about Twitter as a place to have conversations in public, but we actually, historically, have sort of taken our eye off the ball on making conversations better on the service. Just to give you sort of one example on this, if you think about the types of conversations that we support on Twitter today, we have sort of, on the spectrum of conversation, we have tweets, which are on the public record. Anyone can tweet. Anyone can see those tweets. They last forever. They're subject to the popularity contest, that is, likes and retweets. And they're subject to- Public scrutiny with anyone being able to respond. Then on the other side of the spectrum, you have DMs, which are private one-to-one or groups.
28:37 Speaker 05 Maybe in between, you have protected accounts, which is a sort of an interesting use case in and of itself. But for the most part, this, what I believe is a very rich spectrum in between those extremes, we don't support today. For example, one sort of mode of conversation is what we're doing right now. We're having a public conversation, but there's four of us on stage, which means we can have a more controlled, safe conversation where all of the lovely folks in the audience can't come yell in our ears while we're having a conversation.
29:05 Speaker 02 That was the conversation I had with Jack online, but go ahead.
29:07 Speaker 05 Yes, that tends to happen on Twitter, as you experienced with the hashtag #careofjack. That is fundamentally, I think, part of the existential crisis that we have. If we want Twitter to be serving public conversation for billions of people every day, we have to make it not be a cacophony. And I think that inherent in that desire to fix that, there are solutions like adding new modes of conversation that give people a little bit more control around who's participating in the conversation or how long it lasts. And I think we have not explored sufficiently solutions like that in the past, and that's very much part of our strategy now. And I think if we do that, it's one of the many things that can help ultimately deliver more value for people and incentivize them to want to contribute to the public conversation,
29:51 Speaker 05 come back every single day, and so on and so forth.
29:53 Speaker 02 Bijan, one of the things you mentioned, we've got to get to questions. You talked about the idea of white supremacists. I know you've been examining what more to do. I think you guys announced the idea of what more to get rid of them. Is that correct?
30:06 Speaker 03 So we're trying to get a better sense. So we've been working with researchers and academics for many years on how we develop our policies. But obviously, this has hit an inflection point here with some recent attacks that we've seen, Christchurch attack in New Zealand specifically. And we're trying to understand whether there's more we can and should be doing. There probably is. And one of the things that I always like to emphasize is our rules are a living document. They are going to change with the times. They change all the time. We are literally making updates. And we need to do that because we're seeing new harms. We're seeing new ways that people are using platforms to radicalize. And so we've convened a group of researchers to basically continue advising us on whether there's more that we should be doing. As I mentioned, we have a number of policies in place that address these violent extremist groups.
30:49 Speaker 03 Is there more that we should be understanding there and more that our policies and our product can be contributing? You could make the platform not open.
30:56 Speaker 02 This is Peter's thing.
30:57 Speaker 03 That's my thing.
30:58 Speaker 02 Not open or not public?
31:00 Speaker 04 You could just say, "Look, you need to tell us you're not a white supremacist, or other terrible person," before you can start tweeting.
31:07 Speaker 03 So, I think that that's interesting. There's two things I will point out. Please fill out this Google form. Please fill out this Google form. There's two things I'll point out. Number one, I don't think anyone can convince me that bad things don't happen on private platforms. Bad things probably happen more often on private than public platforms. There is an advantage to being open, which is that everyone can see it and respond to it and understand what's happening. Now, there's a disadvantage to that as well, obviously, because you give people a platform. So you have to find the right balance of being able to provide a space for ideas to happen, but counter-narratives to flourish. So, yes, you could do that. But I also try not to view this lens as one problem. This is not just about radicalization in America or in Western countries.
31:52 Speaker 03 This is about what's going on all around the world. 80% of the people that use Twitter are outside of the United States. Most of them don't necessarily engage in conversations around news or politics. A lot of them are just talking about K-pop, quite frankly. So, it's hard to... That's K-pop, and I cave on...
32:08 Speaker 02 Yes, I know that. I know who they are.
32:09 Speaker 03 Yes.
32:10 Speaker 02 It's a band called Carrot.
32:11 Speaker 03 Who they are.
32:11 Speaker 02 Yes, I know the whole genre. But when you...
32:14 Speaker 03 But can I just finish to Peter's point? There are a lot of people in this world, and I spend a fair amount of my time talking to human rights activists around the world, and they greatly rely on this platform to document what's happening. So, it's easy to say, "Let's just change this feature," but this is a product that is used by a lot of people for a lot of very, very important things in the world.
32:34 Speaker 02 But, I'm going to ask a question. But getting back to Trump, your platform has been essentially hijacked by George Conway, Donald Trump, and AOC, as far as I can tell. You know, on some level. You talked about not kicking him off. He's violated things that are in your things many, many times, and you all decided he was newsworthy. Just the way, for example, Robert Mueller couldn't pursue anything because he's the president. When he is not the president, what do you do?
33:02 Speaker 03 Well, we've talked about the fact that world leaders
33:07 Speaker 03 have an outsized influence on the platform. And so we do have a policy that thinks about the newsworthiness of the content. And for a couple of reasons. This content is available in so many different places. Rarely would a world leader say something that's only available on our platform. Oh, but he's used Twitter in a whole new way. There are plenty of people who use whose statements are also covered on media. Even if we deleted a tweet, that would get attention as well. That content would get attention. So I think that the improvement that we need to make here, which we're working on and we will definitely be delivering very shortly, is twofold. Being very, very clear about what's in the public interest. And the balance that we're trying to strike between that public interest of people being able to view and respond to that content
33:51 Speaker 03 and the harm that that content could possibly have if it stays up on the platform. And the second is making it very clear in the product when we're actually making that call. Because it's not fair that this content, if it's a violation of our rules, that would be out there like every other piece of content and have no sort of information around it. So we have work to do there. It's actually going to be coming out very soon. We're very excited about this. But part of this is just transparency. I want, I know that people aren't always going to agree with what we're doing and our policies and how we're enforcing them, but they should at least understand them. Yeah. I'm glad you didn't answer my question.
34:24 Speaker 02 Have a healthy conversation.
34:25 Speaker 03 Okay.
34:26 Speaker 02 We're going to take another break now, but we'll be back after this with Vijay Gade and Kavon Beckpor of Twitter, live on stage at the Code Conference.
34:37 Speaker 06 It's me, the
34:40 Speaker 06 We don't need the jingle twice.
34:43 Speaker 06 Stop it. This is about vacation inflation and how Priceline negotiates amazing deals on hotels, flights, and rental cars.
34:51 Speaker 06 Yeah, but I didn't mention that you can save up to 60% off hotels in the Priceline app. Fine. No one deals more deals than Please stop. Touche. Priceline.
35:05 Speaker 07 How about a latte?
35:06 Speaker 01 Maybe dark roast?
35:07 Speaker 07 Vanilla?
35:08 Speaker 01 Caramel?
35:09 Speaker 07 Blueberry?
35:10 Speaker 01 You could get them to agree on one flavor or give everyone what they want. With Keurig, there's over 500 K-Cup pie varieties of fresh, flavorful coffee.
35:22 Speaker 01 No effort, no grind, no settling. Great coffee without the grind. Keurig.
35:32 Speaker 08 Inaffraid with Axios, how big a problem is the false equivalency issue, particularly when it comes to how women, people of color, LGBTQ people, trans people are treated on Twitter? It seems that the case is that, you know, someone will be harassed. If somebody complains, the person who is doing the harassing says, "My free speech rights are being taken away." And they're more likely on Twitter, someone's more likely to be banned for the harassment they described than the person doing the harassment. And what are you guys going to do about the sort of both sides-ism, you know, conservatives are being silenced notion?
36:10 Speaker 03 So I think one of the things that we're focused on, as Kayvon mentioned, is being more proactive on the health side, particularly being able to detect abuse and harassment. Today, we rely so much on people to report it that I think what ends up happening is if you have a certain group of people who are more likely to either report or not report content, you're going to see some apparent bias on what that looks like. And so if we can get better at proactively identifying this when it's happening and addressing it before it's causing harm, then hopefully it reduces some of this appearance of conflict or appearance of bias. But I'm not sure what else your question is asking.
36:51 Speaker 08 So right now, I'm most familiar with what happens in the trans community. Someone will be harassed, they'll say, "This person's harassing me." The person reporting the harassment is actually more likely to encounter negative consequences than the person doing the harassment.
37:05 Speaker 03 In terms of the blowback effect for the fact that they reported the person?
37:08 Speaker 08 In terms of Twitter taking action. You can't use the word transgender excluding radical feminist, but you can insult a trans person for being trans in terms of actual enforcement.
37:19 Speaker 03 I'd have to look at the specifics. What I will say is that context matters, and I'm sure you've heard this many times on this stage. Content moderation at scale is very, very challenging. There will always be mistakes, and especially as we move to a more technology-focused product solution, we're going to see a lot more mistakes, and we're going to have to be... Very comfortable with appeals and getting things right after the fact. Because it's just the reality of the world that we're moving towards with content moderation.
37:45 Speaker 05 The one thing I would just add to that I think sort of touches on our previous conversation around more holistically addressing this, not just from a policy and enforcement standpoint, but also from a product standpoint, is one of the things we are really focused on is giving customers more control around how they can feel more safe on the platform. And specifically around conversations, the balance of power right now between people who start tweets, people who participate in those threads, and just general sort of people in the audience is sort of is off kilter. And so what we want to do specifically for folks who start conversations is give them more power to do a little bit of the moderation themselves. So one of the specific features that we're going to be experimenting with very shortly,
38:29 Speaker 05 you might have seen some folks tweet about it already, including today, is actually giving authors, people who start a tweet, the ability to moderate the replies within their conversation thread. Now, we think that the author should have a little bit more control, but we also want to balance that with transparency, because you can imagine all the unforeseen circumstances of political authority moderating sort of dissenting speech, and we want to counter that by also having some element of transparency as well. But that's just one example of something that we'll do, of many other solutions around giving people more control around the experience themselves, so that we're not doing everything through algorithms and policy and enforcement. John.
39:07 Speaker 09 Vijaya, when I was growing up, there was no mainstream platform for white supremacists, Nation of Islam, anti-Semitic content. And it seems to me that in the age of social, one has been created. You used to have to go to Idaho to get radicalized, but now you don't. You just have to open up an app.
39:25 Speaker 03 Just Idaho?
39:27 Speaker 09 Idaho was a very popular place to get radicalized, from what I understand. Have you guys thought about, for all the positive impacts, you know, in other countries, perhaps, for people who didn't have a voice, that social, that Twitter has, the negative impacts of radicalization, and to what extent you guys are enabling that potential that didn't exist before?
39:52 Speaker 03 100%, we worry about that. I grew up, I'm a first-generation immigrant. I grew up in Texas, on the border of Louisiana, in a very small town. This was the life that I experienced. This was what my parents had to deal with. And I am very, very focused on that. It keeps me up at night worrying about that, which is why we focus so much on the policies we've had in place around the violent extremist groups, which groups we've designated there, such as, as I mentioned earlier, the KKK, the American Nazi Party. If you have any affiliation, you claim any affiliation to those parties, you are not allowed on Twitter, period. You can't have any accounts. So I want to be very clear that that is our policy. We'll continue to enforce that.
40:37 Speaker 03 We do have work to do in terms of understanding what more we should be doing. And that is the work that we're engaging in. I don't want to make that decision all by myself because there are a lot of experts who work on radicalization on the ground in these communities, engage in these conversations. And I want the benefit of their expertise and their opinions before I make further changes. But those changes are coming. The rules are going to be updated all the time to address new and emerging threats, and this is certainly one of them.
41:03 Speaker 10 Brooke. Hi. I'm at Brooke on Twitter. And a little bit to Ina's point, but I can tweet something about Trump or tweet something about pride or women's rights or whatnot, and there'll be a litany of, I mean, really awful things that people say, really, really abusive. I'm verified. I know Kara Swisher and Peter Kafka. I have connections, and yet I'll complain or I'll put a report in, and within minutes, if sometimes it comes back and says, "Sorry, we've looked at it, we appreciate it," but nothing. I have reported over 500 times. Not once has it ever been solved. So what about the kids, the 17-year-olds, who don't have the access, who aren't verified, who don't have the network, and they don't have a support?
41:48 Speaker 10 You guys have said for years that this is getting solved or trying to be solved, but I don't see it, and it's only increased more and more. And somebody like me, an adult, I'm a big girl, I can take it, but some of these people can't. And the abuse just rises and rises, and I'm worried about people's mental health. I'm worried about what that does to the younger generation, where this abuse is just coming at them and it's not being taken care of by the adult supervision. So is that really a priority?
42:19 Speaker 03 I think one of the things I'm most excited about in our efforts is to switch to proactive enforcement. I think that for too long, we relied on people reporting things to us. And I'm sorry about your experience and the reports that you get. I can't address those specific ones. But I do think enabling Twitter to be more proactive and actually identify this content to the extent that we can is going to make a big difference for people so that we can take action on this content before it even gets seen. So that's really where a lot of my hope is. And I do worry very much about silencing voices. I worry very much about the fact that abuse is directed towards marginalized groups, that it's directed towards women online.
43:03 Speaker 03 That has always been the case, and it continues to be the case. And how can you really have a global public conversation if you don't have those voices feeling safe and enabled to participate? So that is something that we're very focused on. And I think proactive enforcement will make a big difference there.
43:17 Speaker 02 Last question.
43:19 Speaker 11 Thank you. My name is Heidi Steinberg, and my question is, when you have a very high-profile voice on your system that, let's say, you can't take off, what leverage do you have in managing the content and misinformation that may come on? What can you do to influence that?
43:41 Speaker 02 Like Mars and the moon, for example.
43:44 Speaker 03 I'm sorry?
43:45 Speaker 02 Mars and the moon. You didn't see that one?
43:47 Speaker 03 I didn't.
43:48 Speaker 02 Oh, sorry. It was Trump. He said Mars and the moon are related in some way I don't understand.
43:52 Speaker 04 Terrorists.
43:55 Speaker 03 Terrorists.
43:58 Speaker 04 Terrorists. Anything that he thinks is wrong.
44:01 Speaker 03 Yeah, I hear terrorists when you say terrorists. That probably tells you a little bit about my job.
44:05 Speaker 02 Right, okay.
44:07 Speaker 03 So I think, first, I just want to address that there is nobody who can't be kicked off of Twitter. There is no blanket person that gets to stay on Twitter no matter what. And we have spent a lot of time thinking about different options around amplification in particular around that. And again, highlighting credible sources, highlighting conversations that are relevant to the topic, and making sure that there is a balanced perspective brought to the table. I don't know if you want to talk a little bit more about some of the work there.
44:39 Speaker 05 I mean, I think you've covered it, but I think that in general, one of the things that we're shifting more of our attention on the product and engineering side is making sure we're bearing responsibility for where we are amplifying content, because I think speech is one vector, but amplification is the vector that I think increasingly is going to be more important for platforms like us to be responsible for, beyond just whether the content is or isn't up. I think over time, the proliferation of content is going to be pervasive enough that that won't matter as much as what we choose to amplify, what we don't amplify, when we amplify it, what context that we put around it, be that editorial context from us, or editorial context from partners. I think that is an area that
45:22 Speaker 05 is really important for us to keep making progress on.
45:25 Speaker 02 Is that what conservatives call shadow banning, though? Is that the same thing?
45:29 Speaker 05 I think people use shadow banning very... it's sort of a loaded term.
45:35 Speaker 02 Yeah, a little bit.
45:37 Speaker 05 But I think that for us, one of the core principles around anything that we do on the amplification front is to be transparent about it. I think the notion of shadow banning implies that you're not being transparent, because Peter thinks he tweets something and it doesn't actually go anywhere.
45:52 Speaker 04 Again, like we were talking in the beginning, you do want to put your thumb on the scale and say, "This is information we think should be pushed out, this should be amplified, we think anti-vaccination information should not be," and you're comfortable saying that. You can say that a little bit more louder, though, right?
46:06 Speaker 05 Yeah, so long as we're transparent about it. Like, in the case of the vaccine circumstance is extremely transparent, because you search for a hashtag, and boom, we have basically context that's directing your attention towards more credible information. So I think taking that principle of transparency and applying it across different circumstances is really important.
46:28 Speaker 02 Okay, very last question I'll give you. What is the most important product innovation you're making this year? You're going to come on my live Twitter to do the guest thing, but edit button?
46:40 Speaker 05 Would you call that the most important?
46:41 Speaker 02 I do. Everybody thinks we should have an edit button.
46:43 Speaker 05 Casey.
46:44 Speaker 02 No, it's not just me and Casey. It's a lot of us.
46:47 Speaker 05 I would highlight two things personally that I'm most excited about.
46:49 Speaker 02 Okay.
46:50 Speaker 05 One is the work I just described around conversations, around filling out that spectrum in between. The second is making Twitter a better place to find, discover, and talk about your interests. If you think about Twitter right now, it's entirely predicated around following people, which is great. 13 years of product evolution has brought us to a great place there, but as an interest network, it's sort of odd that Twitter does not expose interests and topics as primitives that you can follow or that you can mute. That's a really old idea for you guys. Goes way back to the old, old Twitter. Well, we've talked about it that way for a long time, but we've never really gotten far enough into exposing that as a product experience that you can use as fluidly as following or muting an account. And that's something that I think fundamentally has
47:34 Speaker 05 the opportunity of changing how people interact with the service for the better, both in making it more relevant for you and also giving you control around not hearing the things that you don't want to hear about. Sometimes I just want to watch the French Open and not have politics clutter my feed.
47:47 Speaker 02 What? What's that? Last question for you. What's your number one priority this year?
47:50 Speaker 03 My number one priority, it's hard to pick one. I think we're continuing to focus on election integrity, and I think that that's so critical for all of our democracies.
48:00 Speaker 02 Okay. Thank you so much.
48:03 Speaker 05 Thank you.
48:04 Speaker 02 Vidja, Kayvon, you're very game. Thanks again to Vidya Gade and Kayvon Beckpour for joining us onstage at Code, and thank you to Peter Kafka for conducting that interview with me. And of course, thanks to you all for listening. You can follow me on Twitter at Kara Swisher. My executive producer, Eric Anderson, is at Erica America. My producer, Eric Johnson, is at HeyHeyESJ. If you liked this episode, we'd really appreciate it if you shared it with a friend. And make sure to check out our other podcasts, Recode Media, Pivot, and Land of the Giants with Jason Del Rey. Just search for them in your podcasting app of choice. Thanks also to our editor, Joel Roby. Thanks for listening to this episode of Recode Decode. I'll be back here on Monday. Tune in then.
Previous Recode Decode: The Intercept's Mehdi Hasan Next Recode Decode: Amazon Web Services CEO Andy Jassy (Live at Code 2019) Page updated October 26, 2026 · transcript and summary are machine-generated