Nieman Foundation at Harvard
Three summers ago, in my first weeks getting to know the town of Lexington, Mass. as a local news reporter, I had the typical conversations with community members about what I should think about covering and where I should look for story ideas.
Two names came up early in those conversations — not of people, but of online communities: the Lexington Mavens, and the Lexington List.
The first, a Facebook group with about 4,300 members today, describes itself as “a network of female Lexington residents who communicate, share information, are better informed, and up-to-date on everything happening IN and AROUND Lexington” as well as “a way for the women in Lexington to socialize, connect with each other, build friendships, exchange information, give advice, and perhaps, create change.” The second is a smaller, email-based group (with just over 1,800 members). From the topic tags, you’ll see discussions touch on everything from weather (the top hashtag by far), to local government, to schools, to recommendations for household services. Meanwhile, a sub-list with fewer than 500 members, called Lex PolRel, is reserved for discussions of politics and religion, which are not permitted on the broader List. Both online communities have guidelines members must abide by, and require moderator approval to join.
Eli Pariser and Deepti Doshi have been paying attention to how groups like the Mavens and the List, across platforms, are filling a need for local news and civic connection.
Pariser has been thinking about how to create productive human connection on the internet for a long time. Remember Upworthy (“Man raised in tiny Maine town creates positive-news-focused internet behemoth. You’ll never guess what he did next”)? He co-founded that website in 2012 with the hope that it could “play a part in realizing that vision of the internet as something that actually connects us, not just to the ideas that we already knew we were interested in, but to ideas that we didn’t know we were interested in,” as he told CBS in 2014. His 2011 book, The Filter Bubble, anticipated much of the hyperpersonalized, siloed internet hellscape that defines our online lives today (his Ted talk the same year warned of “a passing of the torch from human gatekeepers to algorithmic ones”).
In an interview last weekend with The New York Times, political scientist Robert Putnam (of Bowling Alone fame) said that alongside in-person membership in community groups, he believes social media could play a productive role in rebuilding America’s disintegrating civic fabric. “We could develop social media in a way that they would actually contribute to our lives — our personal lives, and our collective lives,” he said. The catch: it would probably be “a little less profitable.”
Pariser and Doshi have, essentially, the same thesis as Putnam about the possibility and potential of better social media. They have a vision, and millions in philanthropic backing, to reimagine digital public spaces to better connect people and serve the public good. They’ve teamed up to realize that vision through a nonprofit called New_ Public, which Pariser co-founded with UT Austin professor Talia Stroud in 2019. (Doshi, “a community organizer who has been working at the intersection of social change, social media, and leadership development across the private, non-profit and public sectors,” joined Pariser in 2022 as co-director after seven years at Facebook. Stroud remains on New_ Public’s board.)
Upworthy was a “‘work within the existing system’ project, whereas New_ Public is more of a ‘we need some new systems’ project,” Pariser reflected in a follow-up email.
Upworthy was a venture-capital funded startup “that explicitly intended to bend the algorithmic logic of Facebook and other platforms toward socially responsible things.” Pariser is proud of much of Upworthy’s work, he added, “but at the end of the day our impact was constrained by algorithmic choices that prioritized the company’s engagement metrics and the limits of what is possible in a VC backed business.” Now, “New_ Public picks up where that leaves off, in my mind, by attempting to meet some of the same goals — a healthy digital public sphere — through more fundamentally public-spirited ends.”
Doshi and Pariser have zeroed in on what they see as some of the most civically productive, practically useful pockets scattered across the various platforms constituting social media today: community-oriented ad-hoc groups just like the Lexington Mavens and List. Last month, New_ Public announced a new “Local Lab” that will tackle “a series of research sprints and pilots” focused on talking to, and learning from, moderators of groups like these. Pariser and Doshi hope to ultimately figure out concrete ways to make the ad-hoc groups filling a local news need in communities across the country more resilient and sustainable.
As part of that work, they kicked off an eight-week virtual “Neighborhood Steward Fellowship,” a program “to see if bringing stewards together to share and experiment with new best practices can help improve local cohesion, civic engagement, and social trust.” The program’s stated components include co-creating a best practices guide, weekly Zoom meetings among the cohort to share ideas and get mutual support, and a $1,000 stipend for each participating community steward, with no use restrictions (“keep it as a token of gratitude for your public service, donate it to a local charity, or maybe even throw a neighborhood block party”).
“We’re working with eight stewards/communities right now, across two pilots,” Pariser told me. One pilot “focuses on relationship building between stewards to build social trust in their communities, and one…looks at how stewards might ‘remix’ the information in their communities in a weekly update.”
These groups are “the place, I think, where local news gets discussed and surfaced…either well or poorly,” Pariser said. “We talk to a lot of towns where there is no newspaper anymore; there’s no community center anymore; the town store shut down. And this is kind of it.”
In the short term, Pariser and Doshi are working on existing platforms, meeting community groups where they’re at; groups participating in the pilot operate on Facebook, email, and other platforms. But eventually, they’re interested in new technological infrastructure — whether by connecting existing platforms, or building something entirely new.
A key reason Pariser and Doshi wanted to work with local online groups: They both believe limited scale is a prerequisite for healthy online discourse.
“I fundamentally just don’t think it is possible to do good content moderation or conversation at a scale of billions of people,” Pariser said.
“People referred to Twitter as like a global town square, but it is, in fact, an oxymoron,” Doshi added. “A town square operates at a certain size, not at a global scale.”
In New_ Public’s view, spaces for productive, civically valuable online conversations should be “more like an actual town square in an actual town,” Pariser said, “than this tortured metaphor of a global town square.”
The idea for New_ Public was born (like more than a few save-democracy-oriented efforts) in the wake of the 2016 election. The concept of filter bubbles and the siloed internet was front and center again, and Pariser “was feeling frustrated with the way that conversation was shaping up, which was…very much about, how should we tweak or change Facebook to make it a little bit better?”
To him, that question was too small: “To me, the big question was: What is the information and conversational infrastructure that a democracy needs? And how do we work backwards from that, and evaluate what we have now in light of what we actually need?”
Stroud was thinking about some similar questions at the time, Pariser said, so they teamed up to research “the dimensions of healthy digital environments,” with a focus on how to think about online platforms as physical spaces, and how to structure those spaces in ways that benefit the public good and nurture healthy relationships.
Pariser has described how three key components of healthy physical public spaces have analogues in the digital world: the built environment corresponds to code; programming (what people are doing in the physical, or online, space); and “mayoralty” corresponds to human moderation. Platforms, Pariser said, tend to focus more on code than on the two “softer, social areas.”
What’s more, Pariser told me, he and Stroud were thinking about “the notion that our physical communities have public spaces that are actually built for, and governed by, and serve the public, and our digital spaces are mostly happening in private companies.”
By 2021, Pariser and Stroud were ready to move from research to working on practical applications for improving and reimagining digital public space. Doshi joined New_ Public in 2022 as part of that shift toward a practical focus. She came from Meta, where she’d spent much of her tenure leading the company’s Community Partnerships — heading a team of 60 people worldwide working closely with, and supporting, Facebook group admins.
While Doshi was at Facebook, in 2017, the company changed its stated mission, saying it would work to “develop the social infrastructure to give people the power to build a global community that works for all of us.” Doshi led a department piloting support for community moderators of Facebook groups. The problem was she kept running up against Facebook’s incentive structure, where the customers “are advertisers, not people,” she told me.
Admins would tell her team that they were spending 20 to 60 hours per week managing their groups; they wanted things like team admin tools. Doshi’s team brought those requests to the Facebook product and engineering teams, but found they were “up against other priorities that were more directly related to being able to push growth or engagement for the company.” Ultimately, she decided Facebook wasn’t the place where she could really “build these spaces that are truly serving the public.”
Doshi contrasted Facebook’s incentive structure with that of public media. New_ Public currently collaborates with four public media groups internationally, in Canada, Belgium, Germany, and Switzerland, on a Public Spaces Incubator — another branch of its research and work experimenting with and prototyping better digital public spaces, distinct from the Local Lab.
“When you talk to the director generals of [these organizations]…they see themselves as kind of guardians and vanguards of their democracies. And that’s a really different ecosystem to be operating in than the one I was in,” Doshi said. But these organizations also have “scale and distribution,” which is what originally appealed to her about Facebook in its potential for real impact.
This fiscal year, New_ Public is targeting an annual budget between $5 and $6 million, up from about $4 million in Fiscal Year 2023.1 Its biggest funders in FY 2023 were the Walmart Foundation, Rockefeller Brothers Fund, Jennifer and Jonathan Allan Soros Foundation, Henry Luce Foundation, and the Open Society Foundations. (The Knight Foundation is also a funder.)
New_ Public needs this scale of funding to compete with tech companies for talent “that understands consumer product design,” Doshi noted.
“In order to build local digital space that people actually use,” she said, “we need the same kind of technical and design talent that is attracted to the big consumer tech platforms who also share our moral and public imagination and commitment to healthier spaces!”
“Our budget is high…[relative] to other nonprofits with 20, 25 members on staff,” Doshi added. (New_ Public has 18 full-time employees, operations coordinator Lirael O’Neill confirmed.) “And it is fundamentally because we believe we need engineer, design, and research talent that is comparable and can compete with building and creating things that people want to use.”
The Local Lab pilots with existing civic groups include moderators in “red and blue and purple areas,” Pariser told me, in communities across the country, on different platforms. They comprise neighborhood associations, homeowner associations, town spaces, and other kinds of groups, Doshi added. For the time being, the groups involved in the pilot aren’t public.
“We’re still very much in the learning phase,” Pariser said. “Part of what we’re trying to do is just be in touch with the people who are doing this well, and whose scripts haven’t fallen apart, and who have managed to work through conflict, and learn with them about what’s working.” They also hope to codify some best practices to minimize group shutdowns, he added. Moderators tend to be volunteers, and if conversations veer into toxic territory (national politics being one of the most common pitfalls), that can be all it takes to destroy a rare civil space for conversation about local news.
An early takeaway from their work on this pilot reaffirms something Doshi found in her time at Facebook: Moderators across groups can provide support to each other. “Just by putting them into relationship with other admins, you break that sense of loneliness,” Doshi said. “There isn’t a website to go to, or a certification to go get, to feel confident that you’re trained in being able to handle it.”
Many of these group moderators, Doshi and Pariser said, have no previous experience with moderation — they’re volunteers, not professionals.
“Most of the stewards we meet are people who are trying to do something great with their community [but] have no previous experience facilitating or moderating, dealing with conflict online, [or] dealing with disinformation,” Pariser said. They’re “trying to make it up on their own as they go along, or make it up with a small group of people,” often while also working unrelated full-time jobs. “What we want to do is simply be connecting those people to other people who are having some of the same issues.”
Moderators of most of these groups also work for free, Doshi stressed — putting in a ton of effort, and stepping in to manage stressful conflicts, without any pay, recognition, or support. She’s interested in “innovative forms of compensation” (she called the $1,000 stipend component of the pilot “a minimum” to compensate stewards for their time), as well as development of some kind of “recognition and certification” that would professionalize and legitimize moderator roles.
“In the long term,” she added, “I think we need to develop sustainable compensation models that maintain the public spirited values and culture of their work and acknowledge the time investment stewardship requires. This is critical to ensure that the opportunity to steward is available to everyone, not just to those with disposable income.”
Pariser and Doshi think that shaping conversations from the outset, rather than policing when things get bad, is another cornerstone of successful content moderation.
“This focus on pro-social moderation, where the moderator is actively starting a conversation about ‘Who do you want to shout out who’s done a great job as a neighbor this week,’ or whatever — those kinds of things really work,” Pariser said, “but they take labor that is otherwise allocated to dealing with trolls, or spam or figuring out if someone’s actually living in the town. And so I think we’re excited to also figure out, can we support people with some of the best practices there?”
Shared expectations, or ground rules, can also help bolster these groups and get ahead of conflict.
“Another really big piece of this is establishing norms from the outset that people can agree to, and that you can refer back to,” Pariser added. “Because one of the ways I think these things often blow up is that there wasn’t a clear articulation of where the line was or what the boundaries were, and then someone gets kicked out, and then that person wages a campaign.”
(Unsurprisingly, a common ground rule for maintaining civility is keeping national politics out of local discourse. While this can be helpful, it still creates questions about where the line is between local and national, Pariser noted.)
“We hear a lot of stories about groups…[where] there was one group and now there’s the Biden town group and the Trump town group effectively,” Pariser said. “That just breaks my heart at some level — literally taking this one town space, and fracturing it, and setting those two things at odds.”
“Discourse about anything in America right now is hard, and painful sometimes, and requires a lot of effort to get through,” Pariser acknowledged. On the other hand, for any group where national political toxicity bulldozes a group, “what a loss…to have this network that was so productive and helping people stay informed and helping people trade babysitters, and whatever — all the things that people do in these local spaces — and then have it suddenly go away.”
Since moderating these groups can be such a time suck, Doshi and Pariser are thinking about ways to harness a tool that might not sound like it belongs in a conversation about civic connection and healthy online discourse: generative AI. (They’re not alone — Facebook is piloting AI tools for its group admins, too.)
“AI can play a really important role in reducing this care labor that these stewards are putting in,” Doshi said. The thinking is that AI tools can take on more mundane, time-consuming tasks like (in most cases) verifying whether a prospective member of a given group is actually a member of a given community. It can help take over work where there isn’t a difficult, human judgment call needed — including by acting as a kind of filter to sort out comments that are clearly in-line or out-of-line, and flagging the ones that are edge cases for human moderation.
Doshi pointed to one example of successful self-moderation in a Facebook group with nearly 2 million members. The group, established by Lola Omolola in 2015, was originally called Female in Nigeria (FIN) (now, to reflect its more international focus on women’s issues, it’s called Female IN). The group’s moderators invented a “grumpy cat,” Doshi said, to help gently set and sustain the tone and culture of the group. If someone posts something at risk of breaking the group’s guidelines, users can post the grumpy cat as a shared tool for keeping the conversation civil.
“The grumpy cat is a tool that the moderators use to say ‘this is not exactly what this group is for,’” Doshi explained. “I think that this is an example of where AI can be trained on this kind of spectrum of norms, and then take that work off of the individuals to have to do it.”
For now, Pariser and Doshi are working on existing platforms. But longer term, they want to think about a tech that bridges existing platforms, or creates a new platform entirely.
Different demographics, Pariser observed, are largely siloed on different platforms. How to bring the older users of Facebook together with the younger ones on Discord?
“In the spatial metaphor,” he explained, “right now…there’s a park you can only be in with rollerblades and a park that has all stairs. We don’t have spaces that [pull] in all of the groups and the community… How do we actually bring folks together across those divides?”
Federation is one tool that could potentially break some of the silos and give users more control. They see promise in platforms like Mastodon and Bluesky — though Doshi still thinks Mastodon, in particular, has a demographic silo problem. For now, decentralization “isn’t meeting people where they’re at,” she said.
Before figuring out the ideal technical infrastructure, Pariser said, it’s more important to determine what “a public space that’s welcoming to the whole public” actually requires, and how it can be useful to people’s day-to-day lives, addressing “actual unmet needs that people have.”
“So much of civic tech is like, ‘you should’ — we call it broccoli — ‘you should do this, you should participate,’” Doshi added. “We need to make things joyful, delightful — we need to understand how people engage as human beings who are just trying to get through their days and have fun.”
The tech that meets these needs “might be something that bridges across open and closed platforms, it might be something that’s a whole new platform,” Pariser said. “We don’t want to presume what that looks like yet.”
