Prosocial Design After Social Media

In a court filing defending itself against the US Federal Trade Commission's antitrust case, Meta argued it isn't really a social media company anymore. According to numbers cited by Meta in the filing and referenced in the paper “Towards a Post-Social Media Studies”, only 7% of time on Instagram, and 17% on Facebook, goes to content from friends or accounts people follow. The overwhelming remainder goes toward watching videos recommended by AI-powered algorithms created by Meta.

To date, most prosocial design research has been done on what we used to refer to as "global town square" platforms. These were Facebook, Twitter and similar networks where people posted their own content in public, and anyone could reply to anyone. So if most of what people see in their feeds now comes from strangers or is generated by AI, and their real conversations happen in group chats or with chatbots, our field – those of us working to create and test prosocial design solutions – need to ask where our work matters most now, and how it may need to change.

On June 30, Prosocial Design Network (PDN) hosted a community conversation on that question. As pre-reading, we shared recent pieces: danah boyd on how social media became parasocial, Petter Törnberg and Richard Rogers on what's replacing social media, and New_ Public’s vision on where online public spaces will move “After the Feed.” Below we summarised key points from each of the three pieces, followed by our discussion in the (virtual) room.

Background reading

Social media became parasocial media

danah boyd helped establish the term "social network sites" among academics back in 2007, so it carries weight when she argues in Social Media Is Now Parasocial Media, and that the “social” label no longer fits. In 2006, most people who logged in to social networks posted something because they were building a space to hang out with others. In 2026, posting has waned and most people scroll. What they scroll through is mostly polished content from advertisers and creators who've learned to work the algorithm. Your 15 friends might still post, but buried among polished, professionally made content, you may never see it.

The result, boyd argues, is parasocial media. One-sided relationships with people who don't know you and never reciprocate your engagement. Those relationships can be emotionally intense, but they don't build community. They don’t notice your sad Instagram post and call you up when you’re struggling, and you can feel lonely after hours of following other people's drama. She ties this incentivization structure where first, a platform attracts users with a great service. Once they're locked in, it starts abusing them to benefit business customers like advertisers. When the business customers are locked in too, it exploits both groups to appease investors. Finally, the platform dies. boyd puts most big platforms in its third stage, squeezing both users and business customers to please investors.

But people haven't stopped socializing online, they just aren’t doing it on “social media” platforms any longer. They've moved to texting and group chats, services most people don’t think of as social media. For applying prosocial design research, that could be a challenge – most of what we know about effective prosocial design in mass social media could be irrelevant, or even counter-productive, in spaces that have vastly different features and contexts.

What's replacing the feed

Petter Törnberg and Richard Rogers cover even more ground, looking beyond the big platforms to where social media as a whole is heading in Towards a Post-Social Media Studies. They name three forces pulling the old model apart, largely reflecting the trends boyd described.

The first is the algorithm. Feeds are now built around what a model predicts you'll watch based on past behavior, mostly from strangers, instead of who you follow. “TikTok’s AI-driven algorithm is often called the most addictive on social media” says Weekly Silicon Valley. It set the bar, and Instagram, Facebook, YouTube and X have all moved in that direction. The signals platforms used to determine what you’re interested in have subsequently changed. Likes and comments count for less than watch time, completion rate, rewatching and even the half-second you hesitate mid-scroll. The authors call the result "personalized broadcasting."

The second is generative AI. Generative AI models can create content just for you, and fill up your feed at almost no cost, so platforms depend less on people posting. The surreal "Shrimp Jesus" images that spread on Facebook in 2023 were early examples of this. Even though they were obviously fake, platforms had little reason to filter them out because they got so much engagement.

The third is a retreat from publicness. People are posting less in public and moving their social lives to private messaging groups and closed communities. Törnberg’s own data show posting and viewing on Facebook and X fell by nearly half between 2020 and 2024, and a growing share of people are opting out of social media entirely.

In place of social media, they see three things emerging: algorithmic broadcasting platforms like TikTok, semi-private spaces like group chats and invite-only communities, and AI chatbots, which they treat as a new kind of media. More people, they say, now talk to AI chatbots than post on social media. In 2025, 52% of US adults said they used AI chatbots, while in 2023 only 35% said they’d posted on Facebook, and 11% on X.

For researchers who wish to study these spaces, access is an issue. Platforms keep the watch-time data behind broadcast feeds to themselves, and semi-private spaces are closed by design, so the most socially meaningful conversations are the hardest to see. Chatbot conversations are even harder to study. Each one is private between a person and an AI, so there are no public posts for researchers to look at. Törnberg and Rogers suggest going back to slower, smaller methods that suit these spaces better, like ethnography, interviews, looking at how platforms are designed, and setting up test accounts to see what feeds actually show people. 

Trust over reach

New_ Public's After the Feed, by Eli Pariser, Angelica Quicksey and Arnobio Morelix, focuses on the semi-private spaces Törnberg and Rogers identify. New_ Public offers an explanation for  the retreat from publicness, and looks at the smaller spaces where people talk with their communities and networks once they stop posting to social media. 

Their core observation is that the attention economy is giving way to a trust economy. Once followers and reviews can be faked at scale, engagement numbers stop meaning much. They cite a report from Imperva that 51% of internet traffic in 2024 was automated. That number has gone up, and the most recent report shows bots accounted for 53% of all internet traffic in 2025. When you can’t tell who’s real, open platforms stop feeling like places to talk. 58% of people in the Reuters Institute’s 2025 Digital News Report said they worried about telling real from fake online. So people take their real social lives to group chats, Discord servers, WhatsApp groups and private channels, what Maggie Appleton calls the cozy web: places where you know who you’re talking to and shared norms protect you.

LLM-based chatbots and the agents built on them are also replacing some of the reasons people talked to each other in the first place. For example, monthly questions on Stack Overflow, where developers ask questions and other developers answer, have dropped steeply since ChatGPT launched. New_ Public mapped 134 things people use social platforms for and scored how easily AI could replace each one. Learning and deciding (tutorials, quick answers) scored highest. Connecting and belonging scored lowest, which is why they rate relationship-based apps like WhatsApp and Discord as far less exposed than YouTube or TikTok to being replaced by AI chatbots and agents.

An important point from New_ Public’s deck: private spaces are good for trust and bad for bridging across differences, and democracy depends on bridging. Some of what New_ Public proposes for small spaces, like setting norms and rules, already has evidence behind it. Others, like using AI to support stewards, haven’t been tested yet. How small spaces can connect to each other when it matters is still an open question for prosocial design research.

What we know least about: agents

Across all three readings, what we know least about is how agentic AI will affect Prosocial Design. New_ Public calls AI agents the ‘new new gatekeepers’: an agent may summarize or rewrite your friends’ messages before you ever see them. Participants on our call raised specific worries, like what happens when AI-generated content is forwarded again and again through small group chats, and how easily someone could inject agent-written misinformation into a space built on trust. 

Our Community Conversation

Our community conversation was open to all PDN members on our Slack workspace. (Interested in joining the next conversation? Join us on Slack.) Sixteen members showed up on June 30th, 2026 for a call on Zoom where we asked two questions and then listened. The first question was “What did the reading miss?” The second: “Looking at the future landscape of online spaces, is prosocial design research still valuable and – if so – where should it focus.” We summarize what we heard:

What the readings missed

When we asked attendees to our gathering what was missing from the synthesis in the pieces above, they mentioned four gaps:

Who gets left behind. The move away from big platforms to smaller, niche online spaces could be a global, mass phenomenon, but another – perhaps more likely – scenario is that people who have more resources, money and technical skills move to secure small spaces while everyone else stays on platforms that answer to no one. Rural and low-income communities came up, along with much of the global South. 

Small spaces under big platform control. One participant noted that in rural Canada, local Facebook groups are often the only place for local news and connection. Those groups are based on active participation, and their members depend on a platform they have no control over. 

Social gaming. Gen Z and Gen Alpha build lasting relationships through active play on Roblox and MInecraft and coordinate on Discord. Fortnite draws players across many age groups as a place to hang out as much as to compete. Yet gaming barely appears in any of the three readings, despite being an enormous online industry where many people communicate.

Self-governance. Decentralized and peer-to-peer spaces give communities more responsibility for their shared data and moderation. While that could let people choose how their own spaces are moderated, it also demands more free labor from them. For prosocial design, it may move the focus from platform-wide interventions to the tools, defaults and support that community stewards need. 

Other missings mentioned: Nextdoor and LinkedIn, platforms that still prioritize human interaction between real people, didn't clearly fit into the readings' categories. Also missing were ephemeral content shared on platforms like SnapChat and Instagram, often only with close friends or in DMs.  These are small, semi-private spaces within a larger broadcast platform, a sign of a hybrid big/small future. Because the content vanishes, they are also among the hardest spaces for researchers to see.

Where prosocial design research should focus

In spite of the disappearance of large social media platforms, participants in our discussion generally agreed that prosocial design research is still needed; not surprising among members of a prosocial design network. The consensus was also that research is most needed on large platforms, even as they become less social, and across the many smaller digital spaces emerging alongside them. AI chatbot research, they felt, might be best treated as a separate domain for now.

Big platforms that aren't very social

• Our broadcast environment. Big platforms now work less like places to talk with people you know and more like TV or radio, broadcasting content to each of us based on what an algorithm predicts we’ll watch. But even though they are no longer social, participants thought that their impact on individuals and society – and how prosocial design can improve those outcomes – are still worth studying.
• A hybrid big/small future. Big platforms still hold small communities. Fans of a particular football player might have their own group, and that group sits inside a much larger community of football fans in general, with people moving between the two. So instead of a clean split between big platforms and small spaces, we might be looking at a hybrid big/small future, with prosocial design research being relevant in both domains. 
• Different rules for different users. Age verification and kids' accounts, often required by new laws, mean one platform can give people very different experiences depending on age and geography. Prosocial design research can play an important role in sharing insights in designing for different age groups and countries.
• New research challenges. Studying impacts and design solutions on large platforms may, however, become harder given that they are driven by obscure algorithms based on passive viewing that optimize for watch time, so the field needs new measurements for prosocial outcomes from those too. It may even be necessary to build on innovative research approaches such as conducting research on simulated social media platforms or using in-lab techniques such as eye-tracking.

Smaller spaces

• Don’t assume small means healthy. Some participants expect more prosocial behavior to develop in smaller spaces, as people there work out new norms. Others asked how much we should assume that. Small communities feud and can get “too cozy.” Humans have always gathered in small spaces offline, and those have always had misinformation and antisocial behavior too. A small group chat also isn’t the same as the in-person settings where healthy interaction tends to happen, where people share a physical space and can read each other’s reactions. Prosocial design research can play a vital role in helping those smaller spaces design for healthy outcomes. 
• Whose norms count as prosocial? At the same time, what counts as healthy depends on the community and the wider social context, since different people and communities have different needs. A rule like “stay civil” reflects the values of whoever wrote it, and what feels healthy in one community can feel stifling in another. Participants called for research into the assumptions built into the “social” and the “pro,” and the norms we embed when we design. 
• A move to case studies? The move to smaller, curated online spaces tailored to community needs, means that large-N prosocial design studies on platforms may be harder to conduct. Their findings may also transfer less well, since what works on one big platform won’t necessarily work in a small space with its own features and norms.   Instead we may need to rely more on case studies, looking at the unique sets of features that make some more successfully prosocial than others. 
• New metrics. Part of creating innovative spaces with curated groups and tailored conceptions of prosocial means identifying new metrics. Shallow metrics like likes and shares may be irrelevant. A lot of prosocial design research has measured things like civility in public comment threads, but a group chat between friends can be blunt and full of in-jokes and still be good for the relationships in it. Instead the field may need measures that capture meaningful outcomes like relationship building. For example, on Hylo groups organize activity in the world by posting requests for help and offers of help, so a study can count how many requests get met and offers get accepted, and test whether a design solution raises those rates.

AI chatbots

Prosocial design for chatbots could go in two directions.  It could design for healthier individual use or make chatbots that make groups more cohesive. Participants flagged chatbots as just as important as algorithmic broadcasting platforms or semi-private spaces, given their effect on democracy and society, but the sense among the group – for the time being at least – was that prosocial design research should focus on spaces where individuals are interacting.  At the same time, we see how prosocial design research can inform design for healthier chatbot use, which we discussed some of this at our September Prosococial, Prosocial Design for Healthy Chatbot Use, with Becca Ricks, Project Director of the Public Technology Leadership Collaborative at Data & Society. She leads an expert working group mapping how design choices in LLM chatbots affect user well-being. Their research is forthcoming and we’re looking forward to reading it!

Tell us what we missed

If you're studying any of these questions, or you steward a small community and want to test a design solution in it, we'd like to hear from you. Join the conversation in PDN's Slack.

‍

About the Prosocial Design Network

The Prosocial Design Network researches and promotes prosocial design: evidence-based design practices that bring out the best in human nature online. Learn more at prosocialdesign.org.