The Daily - Jack Dorsey on Twitter's Mistakes

Episode Date: August 7, 2020

It’s been four years since the 2016 election laid bare the powerful role that social media companies have come to play in shaping political discourse and beliefs in America.Since then, there have been growing calls to address the spread of polarization and misinformation promoted on such platforms.While Facebook has been slower to acknowledge a need for change, Twitter has embraced the challenge, acknowledging that the company made mistakes in the past. But with three months to go until the 2020 election, these changes have been incremental, and Twitter itself is more popular than ever.Today, Jack Dorsey, Twitter’s C.E.O., discusses the platform’s flaws, its polarizing potential — and his vision for the future.For more information on today’s episode, visit nytimes.com/thedaily Background reading: A 17-year-old in Florida was recently responsible for one of the worst hacking attacks in Twitter’s history — successfully breaching the accounts of some of the world’s most famous people, including Barack Obama, Kanye West and Elon Musk. But did the teenager do the country a favor?Twitter is in hot water with the government for sharing with advertisers phone numbers given to the company for personal security purposes

From The New York Times, I'm Michael Barbaro. This is The Daily. It's been four years since the 2016 election demonstrated the powerful role that social media companies have come to play in shaping political discourse and beliefs in America. Since then, there have been growing calls to address the spread of polarization and misinformation promoted on these platforms. While Facebook has been slower to acknowledge a need for change, Twitter has embraced it and said that they made mistakes. But with three months to go until the 2020 election, these changes have been incremental, while Twitter itself is more popular than ever. Today, a conversation with Twitter's CEO, Jack Dorsey, about whether

those changes will be enough. It's Friday, August 7th. It's Friday, it's me. It's not just a child. So, I don't know, the reminder. The inner child? The innocent child? Yeah. So if you're ready, I think we are ready.

Ready. Okay. Okay. Um, Jack, I'm just going to start with an intentionally provocative question, which is, do you believe that you are one of the most powerful people on Earth right now? No. Not at all. No.

No. And why not? Well, because if it's a reference to the power Twitter has, I think that power is ultimately in the hands of the people that use it every single day. And that's been the thing that is most special about the service is that everything that has made Twitter powerful has come from the people using it. The people really push the direction of where the service goes and what it is and what it wants to be. And our job as a company, my job as an individual at the company, is to be a checkpoint on that. But isn't that in many ways what you're grappling with? The way in which the people, and in a way the unchecked power power of the people has transformed Twitter. And isn't that a lot of what we're seeing from you now, an effort to kind of moderate that, to rein it in, to intervene in some way, to control that power?

And that's what you see in public conversation in the first place. What we're dealing with, though, is people gaming systems, people taking unfair advantage of systems, people setting up accounts in order to manipulate conversation. And that's where we really need to focus our energy is, is that audience unfairly earned? Is it captured in some way that isn't consistent with reality, which would be a flaw in our system? But are we also dealing with people checking not just themselves, but each other? And I didn't intend to go here so quickly, but this idea of cancel culture that is so present in our society right now? Doesn't that have to do with something that doesn't have to do with gaming systems or taking unfair advantage of them, but rather with the incentive structure of Twitter itself? Yeah, I think you're spot on with incentives. You know, I think if we're to do all this over again and rewind the disciplines that

we were lacking in the company in the early days that I wish we would have understood and hired for are like a game theorist to just really understand the ramifications of tiny decisions that we make, such as what happens with retweet versus retweet with comment. And what happens when you put a count next to a like button? What does that mean? So a game theorist, a behavioral economist to help us understand incentives, and then social scientists.

Those are disciplines that we lacked that I think ultimately would have been important helping us think about not just building a product, but building something that people use socially and the ramifications of that. Meaning if people were using this for more and more conversation, for more and more speech, this is less about building a product and more about how people interrelate with one another. How people converse with one another.

But I think, to me, these things tend to be pendulum swings. And while we do see a lot of what you're labeling as cancel culture today, I do think it's important that we continue to allow the space for people to express their past and their history and context. Because I think context matters so much. Because if we can't express that, we can't learn from it, and then we can't really progress or improve as a culture or as individuals either. You're acknowledging that Twitter has become a difficult place for context.

It really depends on what part of Twitter you look at. 80% of Twitter is outside the United States. 80% of Twitter doesn't really concern itself with what you're bringing up right now. In politics and news, Twitter, certainly there's different ways of using it, some of which are great because they hold power to account. Some are not because it doesn't allow for an evolution of an individual or an institution or learning. But there are multiple Twitters happening in parallel all the time. And we in the U.S., and especially in the media,

tend to focus on one small sliver of it that does have real impact in the world. But from a company and service perspective, we have to pay attention to something so much larger. Right. Well, we also have a leader in the United States who engages in that small sliver in a very big way. Yes. And I want to talk about all of this that we're starting to dabble in in much greater depth.

But you seem to be saying that some very essential elements of Twitter from the start may contain flaws, have contained flaws, or turned out to be flaws. But I'd like to go back with you to the very beginning to understand how it is that we got here. And I want to understand what you thought you were making and what you actually have created. So what did you imagine Twitter was going to be when you created it? And for that matter, why did you create it? What was it supposed to do? Well, I think Twitter is unique and certainly to us in that it wasn't something we really invented. It was something we discovered and we kept pulling the thread on it. And in saying that, we didn't really have any specific intent

around what it should be or what it shouldn't be. We saw some opportunity in technology based on all of our backgrounds and experience. And we were also kind of thrust into a world that had, you know, a country that was just getting access to SMS. And we built it to use it. And immediately we felt it was incredible. The moment that we felt there was something there,

at least for us, was that we all went to our various homes or dinners or yoga sessions or whatever it was away from the office when it finally started working after two weeks. And we all were updating each other about what we were doing. And even though we weren't physically present with each other anymore, we felt together. And it was such an amazing feeling knowing that I would be sending an update and potentially it would buzz some of my friends' pockets and they would take it out

and they would understand in that moment immediately what I was going through, what I was thinking was very cool. It was instantaneous and I was brought into that moment, the moment it was happening. Doesn't a text do that same thing? So what felt so exciting about the fact that the reach of this would be unconstrained? Exactly that. The reach was unconstrained. It wasn't me choosing to send an update to these seven people. It was me writing on the wall and people following updates from that wall. And the number of people following those updates could be infinite.

There was a change in the model so that it was a broadcast and anyone could tune in effectively versus me selecting folks to receive a message. So you understood early on that the potential was kind of huge. Well, no, I wouldn't necessarily say that. I would say that it felt amazing and it felt electric

and it felt very. And it felt electric and it felt, you know, very powerful to us. But what really showed us Twitter were the people and how they used it and how they used it completely differently than how we used it. So I want to talk about that. I want to talk about what I see as the beginning of the transformation of Twitter and the act of tweeting. And I see the start of that really as being the Arab Spring. Young people in the Middle East using this new platform to call for change, to document the calls for change, to document government's response to those calls, and ultimately to bring the change, right, and really alter the course of history, because governments during this period literally fell and rose.

And we see similar usage in the United States with Black Lives Matter, and it felt like that development in the usage of Twitter was very much celebrated, right, especially by progressives, and it felt kind of noble. But this also represents, in my recollection, the beginning of Twitter as a pretty active agent in politics. And that brought a lot of changes to the platform, including hyperpolarization. America's political discourse starts to play out on Twitter, and it gets pretty nasty pretty quickly.

There's harassment, there's name-calling, there's threats, there's anti-Semitism, there's racism. And the idea that everybody who disagrees with me or disagrees with you, you know, that they're just an evil idiot, that becomes a pretty powerful sentiment on Twitter, and it becomes a powerful sentiment kind of quickly. And I know that I just ticked through a lot there. So would you agree with that basic depiction of the transformation? Well, I think it's ignoring everything that happens on the internet in parallel as well. Abuse and harassment did not start after this polarization or the political dialogue coming on Twitter. It's been on the internet forever. And certainly folks in

the early days of Twitter experiencing abuse, hate, harassment, it just wasn't acknowledged, unfortunately, by us in the early days enough or by the general population and general media. So the fact that there wasn't acknowledgement or even observation or stories about it doesn't mean that it wasn't happening. It was. observation or stories about it doesn't mean that it wasn't happening. It was. And it just wasn't being made visible enough and acknowledged enough. Are you saying that was a fundamental reality of human interaction, or at least of human interaction online, as opposed to a result of the incentives and the structures of Twitter?

It certainly has always been part of the structure of the internet and some of what it incentivizes as well versus not. And it's not to say that we didn't incentivize different ways or amplifying the behaviors that already existed, but digital communication has always seen this sort of attack and these sort of trends. So I want to actually dig into what we mean when we talk about incentives and structures. When it comes to incentives, what were they, as far as you could tell, on Twitter? And how did they play a role in both the popularity and eventually the trouble that we're talking

about here on the platform? talking about here on the platform? Well, I think choices around showing how many people follow you and that that number was bolded and big in your profile certainly incentivized me to make that number go up.

That's a number that, for whatever reason, this product thinks is important. That inherently incentivizes people to grow that number as quickly as possible. The decisions we made around having a favorite button on a tweet and then shifting that to a like button and that button having a number associated with it. So people wanting or constructing tweets that went viral and spread as quickly as possible

through retweet numbers. The bigger those numbers, the supposedly better it is. And what was the problem with that specifically? What happens when you incentivize likes and retweets? Well, as you know, it can create behaviors that people are writing headlines for people to click so that eventually people see the ads behind that click. And is that really the right intention

versus informing people about what's happening. And we certainly saw a lot of behaviors where people were constructing tweets just to get as much spread as possible. And then we saw even more sophisticated attacks around that where people found out ways to game the systems in order to get more visibility and to get their message higher than someone else.

So yeah, I mean, I think that spread without necessarily substance is an incentive that can be dangerous. Right, so this is coming back to this idea of nuance and whether that's possible. Or now I guess we're talking about whether it's rewarded on Twitter. Yeah, I mean, you know, one of the things that we are experimenting with

is a small little feature where if you retweet an article that you haven't actually even opened to look at, to read, we will give you a notification that says, hey, you haven't actually looked at this. Are you sure you want to spread it? Because that is a vector for information to spread that might be misleading and people to unknowingly participate

in spreading misleading information. I think there are people with intent and then there's a lot of people who are just kind of seeing things and seeing a headline or seeing a particular tweet and saying, oh my God, and then spreading it without knowing what's in it. And that is on us to help. Why not just implement that? That's a pretty interesting idea. I'm trying to imagine what the downside of that would be, of making sure people have actually seen what it is. You're pointing to some of our biggest issues in the dawn of the company. Why not just launch

something? Instead of think deeply about it and see what the ramifications are. And there are certainly positives, so we can imagine, but there's probably some negatives as well. And if everyone has access to this and everyone is using it, how does that change the discourse? And maybe it's entirely positive, but maybe it's not.

Maybe there's some underlying hidden assumptions that we're making that we need to verify or things that we're not seeing, new vectors of attack, new vectors of abuse. So it's stepping back and thinking deeply about every single small action that we're taking and having a hypothesis, like what we have with this particular feature, and then testing it

and seeing how it plays out on a small scale. And then as we gain confidence around it, yes, launch it to everyone. You're trying to learn the lessons of Twitter today, even as you think about changing the consequences of that. The one thing I want our company to be incredible at, the one skill I want us to build is our capacity to learn. It's a cycle of observe, learn, improve.

If we can be incredible at that cycle, I'm confident we'll do the right things no matter what challenges we're facing. If we become too rigid in coming up with an idea like we just discussed and saying, oh, let's just launch it and hope for the best, we're going to become more irrelevant or dangerous. And we just can't afford that. Would you agree, this might seem harsh, that you have not been incredible about that? And maybe not even especially good at it? About learning?

About the pace of observing, learning, changing, improving? I would agree that we haven't been awesome, but I think we're getting better and better every single day. And I think that is on display publicly, especially in this past year, around everything that we've learned and how we have evolved our policies

and evolved our actions and our enforcement. And I would say that the transparency the company has with the world right now is unique and something I'm very, very proud of and goes much farther than most. So let's talk about algorithms for a moment and how you're thinking about those. What about the intentional surfacing by Twitter

of particular types of messages? Messages that tend to be hot, emotional, to draw lots of eyeballs, controversial. Should that have worked differently? Well, it's not an endpoint. So these algorithms are constantly evolving. So it's not a past tense where we can't change things. But a lot of the algorithms are built

on how people engage with the content. And in the simplest form, you know, are people retweeting this tweet? Are people replying to it? Are people liking it? And if you stop there, then you get to a result where, you know, some of the most salacious or controversial tweets will naturally rise to the top because those are the things that people naturally click on or share without thinking about it or reply to. So there has to be some balancing effect to that. It can't just be a pure read of that signal. It has to take in other signals. I think the most important thing... Was it at one point basically a pure read of those signals? I wasn't fully present in all the decisions around how those were constructed originally,

but I'm sure it was because you have to start somewhere and then you have to evolve based on what you learn. I think it does point to a few issues around algorithms that we as a company and also an industry need to solve for. They are way too much of a black box. They are not written in such a way that they express what criteria they're using to make decisions or even can express how they made a particular decision.

And that's important, certainly, when you consider a ranking algorithm and what you see versus what you don't see. So we need to open up and be transparent around how our algorithms work and how they're used. And maybe even enable people to choose their own algorithms to rank the content or to create their own algorithms to rank it. To be that open,

I think would be pretty incredible so that we can all come to better solutions because it affects society in such large ways. We'll be right back. Jack, can you give us a vision for Twitter, for the future of Twitter? Can you describe how its incentives and algorithm in this vision work? I mean, what they reward and what that version of Twitter actually looks like. Help us see it. Well, what I believe we're building is a conversation layer of the internet.

I believe so fundamentally in the promises of the internet, what it enables in our world. And what I think Twitter represents is the conversation layer of that and the public conversation layer in particular. And I think one of the things I get really excited about as I look at the trends of technology are number one, the trends of translation technology and real-time translation. And I think we're moving from a world where a lot of people had to normalize around the top three languages in order to communicate with like Twitter, I can express myself in my own dialect, and anyone in the world can understand it in real time.

But beyond the language thing for a minute, just because I want to be super clear about this, to give this to me in a way that I think a casual listener would understand, if I'm saying to you right now, for me and for many people, Twitter feels like a place where people go to share and to reshare and to comment on emotional and attention-grabbing and often divisive messages. How do I describe that and experience that in three years, in five years? What's fundamentally different about it? Or maybe it's not.

Again, I think that the reason I talked about translation is because you have more voices and you have more people participating. And I think that is important. We don't have enough people participating in this in a way that they're comfortable with. So you want Twitter to be a better reflection of the whole world, of more people.

Absolutely. But do you think that would solve the civility question that we're dealing with? I mean, isn't it quite possible that more... I'm not saying it solves it, but as we have more representation around the world, then it shifts into another issue and another challenge is how do you focus on relevance and what is relevant to me and what's not relevant to me. And that's where the algorithms come in. And relevance can't just be, did I write something that is contentious

or written in a way that is meant to spread, but is it actually viable and relevant to a certain population or a certain aspect of the population? And there are going to be certain conversations that span communities and span nations and span cultures, but the majority of them are not. The majority of them are going to be more localized. So I think, you know, and I don't know what the timeframe is, but you'll see a Twitter that has this blurring of a very localized conversation, whether that locality be cultural or geography-based or topical-based.

That was not the experience of Facebook in, say, Myanmar. I mean, there are places where social media has been expanded and people have deployed those social media services in pretty horrible ways, in ways that have been genocidal. So isn't it possible that what we're talking about here just spreads across the world? Because if the fundamentals don't change about what is shared, why it's shared, what our eyeballs are drawn to, what the kind of fundamental kind of incentives of this system are in a place like Twitter,

then aren't we just about to export it to other places? Isn't more attention on those problems helpful, though? Isn't more acknowledgement on those issues unfolding in real time in the public important so that we can acknowledge it so that more people can try to help solve them? Some people who may not have had access because they didn't understand it in the past now understand it and can jump in, even if they're not particularly in the location. I guess that depends on if you think Twitter, and I'm curious what you think, is a reflection of society or an amplifier or creator of divisiveness and polarization. It's kind of that chicken and egg thing all over again. Yeah, like any tool, it can be both. With all these tools that we build, all technologies,

we start using them in one way and we discover all these problems. And then we address those and we continue that iteration. This is not unique to this time. It goes back throughout our history as a civilization. You can't pick any tool that wasn't used in some way in both a positive and a negative. And the same is true for the tools of the internet.

The reason I'm asking that original question is I want to know how you are going to try to ensure that it's not both, right? How do you ensure before you grow again globally that you've solved kind of the root problem? And what does that algorithm look like? I think that's just the wrong way to think about it, that there's one solution. But I only heard you really say

that the solution was growth in translation. So I really want to be sure I give you a chance and that we understand what you see as, even if it's not just one solution, maybe a very clear, specific set of solutions. It's just a constant, to me it's a constant iteration. It's a constant push to be steps ahead of how people might utilize this in a negative way,

address that, and then see new potential use cases that are negative. If we try to develop a perfect system that solves the problem, quote unquote, we're just going to get it wrong. And it's going to evolve past that. It's shown itself in that way in so many different cases. I wonder if I can just get an example for you. Beyond the experiment you're running on making sure people have opened articles before sharing, how you might keep iterating for good on Twitter? I mean, there's tons of examples, but one example is another experiment that we've run in the past and that we're running, which is for any particular article that is shared. You might see one point of view in this direction,

another point of view that is slightly different, and another point of view that is completely different. Just to show and kind of break through some of the bubbles that we tend to naturally build. Because I think that the hope and the hypothesis is that you might see these different takes and it might incentivize you to really dig deep into the article or actually watch the video that's being shared

so you can have your own informed point of view and share your opinion as well. And the more of that variety and diverse perspective we have, we get to better answers. Of course, one thing you have done that we haven't talked about a lot here is you've applied a layer of fact-checking and flagging of things that are not accurate and may have significant public consequence, especially when it comes to something like the coronavirus. And as we're talking about your evolution from growth to a focus on questions of speech, I wonder, do you care if groups like conservatives in the U.S. feel like you have a bias against them? I mean, one of your peers, Mark Zuckerberg, does seem to care about that quite a bit.

Do you care if some meaningful percentage of Americans feel that you are somehow suppressing them, censoring them as part of your iterations and your growth? I absolutely do care. I come from a very conservative state of Missouri where St. Louis, my hometown, is much more liberal. So it's in this basket of conservatism. And my dad is very much a conservative. So I absolutely do care that we're building a system that does not take our own bias into account, but feels fair.

And I think one way to show that is continue to be a lot more transparent around our decisions, continue to be clear in our policies, which we haven't been in the past, and we haven't been transparent around our actions and the why behind it. I think that brings us naturally to President Trump and his own use of Twitter.

And there's perhaps no one who has better grasped the incentives of Twitter and how to exploit them than Donald Trump. I once sat in his office with him before he was president, and he talked to me about how savvy a user of Twitter he was and is. And he's turned out to be very savvy about it. Would you agree with that? I mean, that Donald Trump is one of the most deft users of this platform you've created? He's definitely used it to great effect. I wouldn't say he's necessarily the most,

because it's just really a question of what your goals are. I would say his usage tends to be consistent. He started in a particular way, and that has remained consistent to today. But I guess what I'm saying is, he's taken advantage of the existing incentive structure. Does that feel right? And he continues to use it to his advantage. And given the emotional quality of his tweets, it suggests that Twitter is still very much rewarding those incentives.

How would you say that? How would you say it's rewarding? Well, if... Versus society rewarding that, versus the media constantly pointing the cameras on that and putting all the attention on that? How is people using Twitter different from the approach that we're seeing elsewhere?

I believe that was a rhetorical question, but TV is a mediated, somewhat mediated form. And for the most part, until very recently, when the president wanted to say something that contained misinformation, that contained an outright lie, that called a judge a name, that expressed a racist sentiment,

that pretty much was his to-do on your platform without any real layer or mediation. And it got tons and tons and tons of retweets and pickup and attention. Well, you're completely ignoring the layer of people who push back on any one particular tweet or reply to it or spread it with a correction to their followers and to hashtags or search terms. So there is mediation, but it's a question of the people doing the mediation versus the centralized media doing it. So you're right there. And again, a lot of our policies and focus in this particular area are really focused on the velocity and the spread of information and the gaming of these systems.

And where this might have harm if people were to see it and take it out of context. So we did label tweets of his that we felt could be harmful because they may have led people to believe that they were registered to vote when in fact they weren't. And we didn't take the tweets down. We appended them with, annotated them with information as to the facts expressed by the various institutions that were doing the work around registering voters.

So I think it's important that we do recognize, number one, that these annotations are happening by the crowd in real time, all the time. And number two, there are particular areas such as voter suppression and election integrity that we should also take action upon. And we should make that policy as tight as possible. And we should make those interventions as infrequent as possible. But the reality is they have to be there. Are you prepared to ban Donald Trump from Twitter if you feel that he has repeatedly violated your

rules, your terms of service? And I ask this because you have banned certain figures from Twitter, the most famous being Alex Jones of Infowars, because you all said that he had violated your rules around abusive behavior. So does Donald Trump break those rules? And, you know, I mean that because, like I said before, he attacks judges, or he calls women dogs, or he spreads false and misleading information on a routine basis. Yeah, so we, independent of any particular account, we hold all accounts to the same rules. But, you know, if there are particular egregious aspects of violations of returns of service, we won't hesitate to take action on the accounts and use every tool that we have together with us.

So independent of the U.S. president or any leader around the world, we will take action if we feel it necessary. I know people resist hypotheticals, especially people in your position, but sometimes they can be important. And let's say that it is November 3rd or 4th or 5th or 6th or 7th, because this is going to be a very unusual election. And President Trump takes to Twitter and declares that he is the winner, even though that's not yet clear or accurate. And in some ways, our democratic system at that moment is going to be very severely tested. And if he makes that declaration on Twitter, what do you do? Well, I guess I would look for opportunities to learn from the past. So do we not see some of this play out way back during Gore and Bush in Florida,

in terms of being confused as to what the end state around what the election was and how that evolved. So look for lessons in history and work with our peers and civil society to really understand what's going on and then make an informed decision. But as you said, it's a hypothetical that we just need to think a lot more about in terms of the integrity of the conversation around the election and what that see us continue to evolve our policy to protect the integrity of the conversation around elections. Listening to you talk, Jack, you are very

measured. You are calm, you are careful, and you're the CEO of a platform that in its current form thrives on emotion. That's notorious for elevating the most, as we've said, kind of hot, sensational, charged views. A colleague of ours, anticipating this conversation and kind of knowing your tone, joked that if you tweet something that sounds like Jack Dorsey, it probably won't do all that well on Twitter. And honestly, Twitter doesn't seem like you exactly, and you don't exactly seem like Twitter. So do you like Twitter right now? Do you like what it's become?

Do you like what it is as it exists today? One view of Twitter, and I think it's a very specific view, is that it's all focused on these reaction know, these reactionary, emotional headline clickbait tweets when that's just not the reality of the majority of our usage in the world. And not that it's not important to focus on news and politics and how that changes the discourse and not that it's not important to help do everything that we can to fix it but i think the the way to do that is is to listen and to use the tool in such a way that like we can really understand how society is evolving how technology is evolving that we can utilize to to help these problems in the in the first place so yeah i'm not i don't i don't use twitter to get as much spread as possible. I use it to listen

and to observe and to understand our world and my world and myself, ultimately. But if Twitter doesn't change meaningfully from its current form, does it remain deeply flawed? It would be silly for us not to change Twitter. So yes, it should become irrelevant if it doesn't change, if it doesn't constantly evolve, and if it doesn't recognize gaps and opportunities to get better. So absolutely, we would earn that irrelevance in that particular case. Well, Jack, thank you very much.

We appreciate your time. Thank you. Thanks for doing it. Thank you so much. We'll be right back. Thank you. Tennis Getter, Paige Cowett, Michael Simon Johnson, Brad Fisher, Larissa Anderson, Wendy Dorr, Chris Wood,

Jessica Chung, Stella Tan, Alexandra Lee Young, Jonathan Wolfe, Lisa Chow, Eric Krupke, Mark George, Luke Vanderploeg,

Kelly Prime, Julia Longoria, Sindhu Yanasambandhan, MJ Davis Lin, Austin Mitchell, Nina Patoria, Sindhu Yanasambandhan, MJ Davis-Lynn, Austin Mitchell, Nina Patuk, Dan Powell, Dave Shaw, Sidney Harper, Daniel Guimet, Hans Butow, Robert Jimison, Mike Benoit, Bianca Gaver, Aastha Chaturvedi, Rochelle Banja, and Liz O'Balin. Our theme music is by Jim Brunberg and Ben Landsberg of Wonderly.

Special thanks to That's it for The Daily. I'm Michael Barbaro. See you on Monday.