The 2016 presidential election was a depressing experience for Laura Carpenter, an engineer and student at Harvard Business School. “I was frustrated by the toxic nature of our political discourse,” she says. “I’m from the South. I have a lot of friends back home who have strong conservative views and I live in one of the most liberal places in the country. So it just bothered me to be in one place or the other and hear stereotypes about the other side.”
A few weeks after the election, Carpenter started to talk to people across the political spectrum about the way they approached the news. It turned out that they weren’t actually that partisan. They wanted to read different views. They just didn’t have the time to find them.
This is the problem Carpenter and her co-founder David Byas-Smith are trying to solve with Abridge News, a startup currently based at the Harvard Innovation Lab.
Polarization is not a new phenomenon in American politics. Historian Richard Hofstadter described its rise in “The Paranoid Style in American Politics,” published by Harper’s in November 1964. The piece explains how different movements have exercised this paranoid style throughout American history by targeting Masons, Catholics, Mormons, or Communist spies. All of those movements shared unrealistic goals, a sinister enemy, and a vision of politics as a moral fight. “Since what is at stake is always a conflict between absolute good and absolute evil,” writes Hofstadter, “what is necessary is not compromise but the will to fight things out to a finish.”
It is easy to read Hofstadter’s words today as a prophecy on the political, racial, and generational divide that exploded in 1968: the assassinations of Martin Luther King and Robert F. Kennedy, the protests against the Vietnam war, the riots in Newark and Detroit. But the “paranoid modes of expression” Hofstadter described remain.
Harvard professor Robert Putnam points to the decline of social connections as a possible factor in present-day polarization, while journalist Bill Bishop argues in “The Big Sort: Why the Clustering of Like-Minded America is Tearing Us Apart” that Americans accentuated the divide by moving to neighborhoods where they live with people with the same ideology, income, and race.
In the last few years, the debate about polarization has turned to social media platforms. Authors like Cass Sunstein, Eli Pariser, and Ethan Zuckerman have written extensively about the way Twitter and Facebook could be shaping political discourse. They argue that personalized feeds create echo chambers where everyone’s views are confirmed and reinforced. In this kind of environment, Sunstein argues, misinformation flourishes, radical arguments are popular, and common experiences are scarce.
Not everyone thinks that the rise of social media platforms is transforming the public sphere in a significant way, however. A working paper published in 2017 by three economists suggests that polarization is accelerating fastest among those using the Internet the least.
Whatever its causes, polarization is a fact of political life. In response, journalists, entrepreneurs, and technologists are trying to transform the media ecosystem by building projects whose goal is to improve civic discourse. Here are four of them.
Abridge News
The Harvard Innovation Labs are a world away from the plush quarters of the Business School’s nearby buildings. Surrounded by free snacks and whiteboard walls, dozens of teams work every day in this open-plan office, where people bring their own laptops and meet with potential investors.
Harvard founded this incubator in 2011 to help its students collaborate and hone their entrepreneurial skills. Abridge News, the startup founded by Laura Carpenter and David Byas-Smith, is one of more than 150 teams working out of the lab.
The goal of Abridge News is not to fight polarization but to create a site where everyone has access to different views. Here’s how it works: Every day a member of the team selects a news topic and scans the web for four op-eds with different views on that topic. Abridge News publishes a summary of those articles and presents them on a spectrum, with the most radical articles at the extremes and two moderate pieces in the middle. It also publishes an introduction explaining why the topic is relevant and what are the key facts you should know about it. Readers can click a button on each article to say if they agree with its arguments. They can also click on the link and read the original piece.
On April 18, for example, Abridge News published a short summary of the tax legislation passed by Congress in December 2017. The summary included a couple of figures from the nonpartisan Tax Policy Center and introduced four articles about the topic. On the extreme sides of the spectrum were a New York Times opinion column by liberal author Timothy Egan and an article by U.S. President Donald Trump. Egan’s piece argued that the legislation was structured to benefit the very wealthy at the expense of the average worker. Trump’s article, published in USA Today, said that the law had been passed “without a single vote of Congressional Democrats.” The more moderate pieces were an op-ed by House Speaker Paul Ryan that mentioned the doubling of the child tax credit and an editorial from The Gainesville (Fla.) Sun warning that the new legislation could create a debt crisis very soon.
Abridge News launched its website in October 2017 and added a daily newsletter a few weeks later. Its readers are mostly young and like to consider every side of an argument. The number of reactions to each article is almost identical, meaning that people usually read all the views about a topic, not just the ones they agree with. The site has a loyal audience of thousands of users and has grown by 30 percent every month since it launched. “Our goal is not to bring people towards the middle but to build empathy and understanding between people with different opinions,” says Carpenter.
Are they succeeding? The early evidence is encouraging.
The team in the spring introduced a new feature that allows anyone to express more nuanced reactions to the views featured on the site. For example, readers who agree are given three additional options: “This makes sense to me,” “Surprisingly, I agree” and “I strongly support this.” Data from this new feature suggests that some people agree with views that are not their own. On tax reform, 26 percent of the people “surprisingly agreed” with the view from the mid-left and 15 percent “surprisingly agreed” with the article by President Trump.
“We would like to double down on the context of the opinions we are featuring by doing two things,” says Carpenter. “Helping readers to understand more basic facts before they engage with the articles and link related topics together so that a reader can see how opinions about a subject like DACA or North Korea evolve over time.”
Tracking their readers’ reactions to different articles is something that could be useful for publishers and authors who want to know how persuasive their arguments are for different audiences. It could also help shape the business model for Abridge News. “Our ambition is to build a platform with millions of users,” says Carpenter. “By capturing their nuanced reactions, we will be able to track how views are evolving over time. This information can provide valuable insights we could sell as polling companies do.”
Right now the team is focused on user growth and not yet on pursuing monetization. “At scale, we will run tests to see which types of business models may make the most sense,” says Carpenter. “We imagine that monetizing through a combination of ads, subscriptions, and data could be effective. But we are very aware of the concerns about Facebook and Cambridge Analytica and will spend a lot of time educating ourselves on data privacy concerns.”
Abridge News is open to featuring radical arguments as long as they don’t peddle inaccurate facts. Every article is curated by a three-person editorial team who ensure they don’t include details that have been debunked. “All of them are opinion or editorial pieces,” says Carpenter. “They may be biased in that they focus on a subset of facts while not addressing the whole picture but these opinions are counterbalanced by others along the spectrum.”
Abridge News doesn’t always feature political topics. They’ve also featured articles about sports, business, and entertainment and published spectrums of opinion where the debate couldn’t be forced into a choice between left or right. The extreme views are defined by two pieces that take the strongest point of view at either side of a debate. Carpenter is fine with polarization as long as people know more about the other side: “The platform is not about making liberals more conservative or conservatives more liberal. I couldn’t care less if our politics stay as extreme as it is today as long as people take the time to empathize.”
Kialo
Errikos Pitsos tracks his love for reasoning to his personal background. His parents are both philosophers, and he grew up in a household where people could disagree and laugh about it. “The most urgent problem we have right now is that we don’t have the means to communicate our different points of view to each other,” says Pitsos. “We are just shouting.”
Platforms like Twitter and Facebook thrive on content that promotes emotional responses. The one Pitsos has created is designed to host thoughtful discussions about any issue. He began to develop the concept in 2012 and launched it in August 2017.
Pitsos called his platform Kialo. The name means “reason” in Esperanto and aspires to describe the rational attitude of its users. The site has hosted thousands of debates in its first few months. This is how it works. Every debate is framed around a proposition and is structured as a decision tree in which each branch is “a claim.” Every claim must be unemotional, shorter than 500 characters, and make an original point about a proposition or about another claim.
Claims can link to external sources. Their author is not immediately visible and they must be submitted before publication: the creator of a debate or the users designated as administrators decide if claims are approved or dismissed. Users can thank other users for their contributions or ask them for a clarification. They can also bookmark a claim, mark it for review, or rate its impact on a 5-point scale.
The user who creates a debate decides whether it’s private or open to the public. Without being overwhelmed by false claims or personal abuse, Kialo has featured popular discussions on topics such as racial profiling, bullfighting, or the future of American football. “We only deal with contentious topics but the debates are not personal,” says Pitsos. “It doesn’t make sense for trolls to be in the platform. Their posts are just ignored and deleted.”
One of the most popular debates at Kialo is the one about gun control. It was created by a user called “PaulMoore” the day after the 2017 Las Vegas mass shooting and it has gathered more than 2,000 claims.
Every debate starts with an introduction that establishes some rules and basic definitions. The one about gun control explains what the phrase means in this context and discourages users from writing any claims related to the Second Amendment. “This discussion assumes that the Second Amendment will be modified or interpreted in such a way as to allow these restrictions,” the introduction says.
At the highest level of the decision tree, you can find claims for and against the proposition. At lower levels, there are claims for and against another claim. At one point this spring, the gun control debate had seven claims for and 10 claims against the proposition. One of the arguments against it, for example, is that any restriction would be harmful to gun manufacturers. Another is that any measure would be very difficult to implement.
Among the arguments supporting the proposition, there’s a link to a Gallup poll and the claim that “polls suggest that most American citizens support stricter gun control legislation.” There were four claims for and 13 against this argument. Some argue the poll is flawed or inaccurate. Another one makes historical analogies: “Appeal to popularity is a logical fallacy. Slavery was once quite popular, as was Adolf Hitler in Germany.”
Since its launch, Kialo has hosted more than 10,000 debates in various languages. About 40 percent of the traffic comes from the United States. The most popular discussions are the ones around politics, ethics, and religion. Two debates on the existence of God have garnered more than 2,000 claims. For the time being, a debate stays open forever but that could change. Kialo doesn’t analyze the interactions of its users. “We prefer to focus our energy on improving the platform based on the user feedback,” says Pitsos.
Kialo is considering empowering its best users to intervene in any debate on the platform and help others write their arguments in a better manner. One of those power users is Rhianna Taylor, a 30-year-old special needs teacher who grew up in California and now lives in Colorado. Taylor discovered the platform in November 2017, right after the mass shooting at a church in Sutherland Springs, Texas. She was tired of reading angry comments about guns on Facebook and noticed the Kialo debate on gun control. “Social media threads are just a shout factory,” says Taylor. “You just say how you feel about an issue and people respond by shouting louder or calling you an idiot. Those threads are filled with emotions. It’s difficult to find a single fact.”
Taylor is used to discussing politics with people who have different views. Her mom and dad vote for different parties and have opposite perspectives on most issues. Taylor is also more liberal than her husband, a gun owner who is in the military and opposes restrictions on guns. “Kialo’s debate on gun control has allowed me to take a step back and understand the arguments of the other side,” says Taylor. “It has made it easier for me to tap into my husband’s mindset and has opened up my ability to talk to him about this issue.”
A few weeks ago, Taylor was made an administrator of the gun control debate, and she has since helped to revise claims. Most of the decisions are clear-cut. When she isn’t sure about one, she asks another administrator. Most of her work has to do with claims that are duplicates of other claims or unrelated to the topic. When users express themselves in a way that is too emotional, she works with them to reframe their arguments in a more reasoned way. “Most of them are not academic writers,” she says. “They are used to talking with their friends and these kinds of debates are different. You can’t write things like: ‘I know many gun owners and they didn’t shoot anybody.’”
The platform has had a positive impact on Taylor’s life. Before Kialo, she would get so frustrated with gun advocates that she couldn’t value their perspective. “Now I have an opinion that is more nuanced and not so strong as the one I had before,” she says. “I used to think, ‘How is your ability to own a high-capacity assault weapon more important than stopping the next group of children being murdered? You get all of your information from Fox and Friends, so I’m not going to try to reason with you.’ After participating in the debate for several months, I ask myself, what kind of legislation is the most plausible compromise? Now I can see the whole discussion with logic and evidence on all sides. There are never just two.”
Kialo doesn’t have any external investors. It is funded by Pitsos, who says it has around 50 full-time employees.
Along the way, Kialo has encountered some intriguing use cases. Harvard professor Dustin Tingley has taken advantage of the platform to organize some debates in his class. A few companies have expressed an interest in using Kialo to improve the way they make difficult decisions or to assess the reasoning ability of a potential employee. Pitsos also thinks Kialo could be a great tool to create experiments of deliberative democracy inside local governments around the world.
Pitsos is still figuring out Kialo’s business model but he doesn’t intend to run any ads. This is the reason his team doesn’t care about real names and encourages people to sign up with a pseudonym. Pitsos explains they intend to make money by selling the platform to companies as a deliberation and decision-making tool.
Several publishers have expressed an interest in integrating Kialo into their editorial product. Some want to offer the platform as a service for their subscribers. Others just want to embed Kialo’s debates in their sites. Pitsos says they are considering allowing partners to build their own portals in Kialo or to run their own white-labelled Kialo portal on their own sites.
Spaceship Media
A few months before the election, journalists Eve Pearlman and Jeremy Hay founded Spaceship Media, a startup devoted to create rich conversations between groups of people with different views. “Our mission is to reduce polarization, build community, and restore trust in the media,” says Pearlman. “Many people are very eager to understand the other side of the argument. We offer a structured environment to do precisely that.”
Spaceship Media’s first project was a series of dialogues between police officers and students of color at a high school in Alameda, California. After the birth of Black Lives Matter, the project tried to build empathy between both groups.
A few months later, Donald Trump was elected. Like so many journalists, Hay and Pearlman wondered how they could do their jobs in a more productive way. In partnership with the Alabama Media Group, they created a conversation between 25 women from Alabama who voted for Trump and 25 women from California who voted for Hillary Clinton. The project gained national attention and created understanding across both groups. A year and a half after it finished, two thirds of those women are still in contact and talking through a Facebook group.
The success of this project produced a playbook that Spaceship Media has replicated and refined in different projects: a conversation about farming in Minnesota, a dialogue about immigration in California, and the project they’ve just launched about guns.
Participants are selected by Spaceship Media with the help of its partners. Nobody is paid to contribute. Pearlman and Hay usually ask every candidate if they are open to talk with people who think differently and remind them that the project is a conversation, not a debate. “We’ve had people from the furthest right of the spectrum to the furthest left, and that’s very encouraging,” says Hay. “Those people may be polarized by the nature of the political climate. But given the opportunity they are still willing to talk.” In the conversation about immigration, there was a young DACA recipient, a Tea Party lawyer, a businessman who crossed the border illegally, and a construction worker who attributes his unemployment to undocumented immigrants.
Spaceship Media’s conversations last around a month and usually take place in a Facebook group. “We make very clear from the beginning that this is not an exercise to change anybody’s mind,” says Hay. “However, we often see nuances emerge during the conversation and people showing compassionate feelings. We see people’s minds growing and opening rather than changing.”
Hay and Pearlman’s role is to ensure every conversation is a meaningful experience. As questions arise, they sometimes answer them with their own reporting. They create what they call fact stacks, short explainers with figures that are relevant to the debate. When people challenge those facts, the journalists explain how they selected the sources. The longer the group goes on, the more trusted the journalists are.
Moderating these conversations is not easy. Participants get angry and emotional when telling their stories. They can debate for hours about what terms to use when talking about an issue. Hay and Pearlman establish a few rules at the beginning. No memes or name calling is allowed. People can’t post a chart or an article without explaining why.
“Moderation is a process that starts with the relationships we build,” says Pearlman. “It’s important to help them feel known and recognized and also identify stereotypes they could have about other people. We’ve deleted comments, closed threads, and asked people to leave. But those events are rare. We want people to be who they are.”
For every project Spaceship Media charges a fee to its media partners. The company has also received grants from several foundations. Hay and Pearlman are now raising funds to widen the reach of The Many, a project they’ve just started to create a national conversation with thousands of women before the mid-term congressional elections.
Spaceship Media measures its impact by the way participants enrich their views about different issues. One example is the project about guns they started a few weeks after the mass shooting at Stoneman Douglas High School in Parkland, Florida.
With the help of several media partners—including Advance Local and Essential Partners, a dialogue facilitation firm—21 people were invited to spend a day and a half at the Newseum in Washington D.C. Among the participants were victims of gun violence and NRA members. They shared personal stories and explained their views about guns. “It was a remarkable gathering,” says Hay. “There were people in tears after the event.” Now those 21 people will join around 130 people in a Facebook group. The organizers hope that the time they spent together in Washington will transform them into a moderating force inside that group.
Cortico
While told to stay indoors for a couple of days after the Boston Marathon bombing in 2013, Deb Roy read a tweet that said a bomb had gone off in Harvard Square. A few hours later, he found out that there was no bomb. The story was one of the many pieces of false information published in the aftermath of the attack.
What happened during those days inspired Roy to study the spread of misinformation and its impact on the public space. For a few years, he worked as chief media scientist at Twitter. Now he is the director of the Laboratory of Social Machines at the MIT Media Lab and the founder of Cortico, a nonprofit devoted to strengthening the American public sphere.
Roy is one of the authors of the largest-ever study on misinformation. Published in Science in March, the study analyzed 126,000 stories tweeted by three million users over the last decade. Some of its findings are striking. False stories are 70 percent more likely to be retweeted than stories that are true. The study analyzed the sentiment expressed by users when dealing with misinformation. False stories usually prompted replies expressing surprise and disgust and those feelings fueled their dissemination across the web. “Lies tend to propagate faster and further,” says Roy. “Falsity dominates truth for this category of contested news.”
The name of the company is a statement. Cortico is a reference to “cortex,” the part of the human brain where senses come together to help us understand the world. Roy thinks that the public sphere is reaching a dangerous point. “Polarization turns out to be a great business model if you are a media company,” he says. “Dividing and enraging audiences allows you to sell engaged audiences to advertisers. Our capacity to trust any facts seems to be slipping away.”
One of the first things Roy and his colleagues have done is develop what they call “health indicators”: four ways to measure the quality of a conversation in social media and beyond. The first two indicators are shared attention and shared reality: whether there is overlap in the topics people are talking about and whether everyone agrees on the basic facts about those topics. The other two indicators are diversity of opinion and receptivity: whether everyone is exposed to different perspectives and whether they are civil, listening, and open to changing their views.
These health indicators could measure the quality of conversations in the public sphere at rancorous times like the 2016 election. “Partisanship, Propaganda, and Disinformation: Online Media and the 2016 U.S. Presidential Election,” a study published in August 2017 by the Berkman Klein Center at Harvard, explained how Trump succeeded in shaping the narrative of the campaign around immigration and Hillary Clinton’s emails. Both Republicans and Democrats focused their attention on those topics but posted different articles from different outlets. People from both sides did not have a shared reality when talking about the campaign. However polarization was asymmetrical: people on the right were more isolated and radicalized than people on the left.
As for the other two indicators, data suggest that a significant portion of voters shared articles from partisan outlets and didn’t accept facts that challenged their views. Those online communities lacked diversity and were not receptive to stories or data that could change their minds.
Cortico staff plan to seek input from journalists, legal scholars, designers, political scientists, and behavioral economists to refine these four indicators. Jack Dorsey, CEO and co-founder of Twitter, saluted Cortico’s idea in a thread and said he wants to work with Cortico to develop Twitter’s own indicators of healthy conversations. Roy says, “There’s a version of this that only Twitter can build because they have access to data we’ll never have. This is true for every other platform out there.”
The ultimate goal of Cortico is to come up with recommendations for platforms, publishers, and advertisers about how to create a less polarized environment. In the meantime, Cortico is working with reporters and editors who want to understand their audiences and make sense of the public sphere. For example, Cortico data journalist John West helped the AP track the declining impact of Trump’s tweets during his first 100 days in office. West used data from Twitter and from the MIT Media Lab to analyze engagement with the president’s official account and sorted them by demographics and ideology.
Cortico is also helping documentary-maker Frontline analyze the way its work is shared across the Internet. “Frontline used to have a politically balanced audience,” says Roy. “This has changed in the era of Facebook and Twitter, and we are working with them to depolarize the reach of their content online.” Cortico is looking at who is propagating Frontline content and using those insights to design experiments to bridge the political divide and then monitor to what degree the reception of that content is civil and tolerant across political boundaries.
Cortico is meeting with advertisers eager to avoid crises like the one that had advertisers fleeing Laura Ingraham’s show on Fox News after she criticized a Parkland school shooting survivor. “There are erratic surprises waiting for every brand unless they have a systematic way to know where the healthy parts of the public sphere are,” says Roy.
Cortico sees an opportunity in an environment where public perceptions are evolving and platforms are under siege. “There is an existential threat on the table,” says Roy. “Founders [like Mark Zuckerberg] are coming to D.C. This is unusual. There’s a public awakening about this. It’s hard to sell someone insurance for an extreme event until they experience it. We are in that mode right now.”
Editor’s note: An earlier version of this story failed to mention Advance Local and Essential Partners as partners working with Spaceship Media for their gun violence project.
