The Next Great Social Network May Be Less Social and More Like a Public Utility

What if the central problem with social media is not that its algorithms are too powerful, but that they are solving the wrong problem?

Most feeds are built around a simple assumption: people want to see what other people are doing. The system gathers posts from friends, ranks them according to predicted engagement, and presents the result as a personalized stream. It is an elegant machine for turning human relationships into attention.

But there is another possibility. A social network could treat each person less as a member of a crowd and more as a query, a moving set of interests, needs, curiosities, and aspirations. It could match that query not only with posts from friends, but with the world’s store of useful ideas, discoveries, tools, and opportunities.

That change sounds technical. It is actually philosophical. It asks whether digital networks should primarily distribute signals of popularity or help distribute the things that make human beings and societies more capable.

The surprising connection is this: the future of social media may depend on the same principle that governs global progress. Ideas, institutions, and technologies are nonrival goods. My learning something does not prevent you from learning it. Brazil adopting an effective public health system does not stop America from doing the same. A better feed, like a better institution, can increase what is available to everyone.

The question is not merely how to build a more pleasant feed. It is how to build an information system that turns attention into a positive sum resource.

The feed is an institution disguised as a product

A feed appears to be a user interface. In practice, it is an institution. It decides which possibilities become visible, which creators receive distribution, which beliefs seem common, and which problems appear urgent. Its ranking function is a form of governance, even when the company describes it as personalization.

The conventional architecture has two stages. First comes candidate generation: from an enormous universe of possible content, the system selects a small set. Historically, much of that candidate pool comes from friends and accounts a person already follows. Second comes ranking: the system orders those candidates, often by estimating which ones will attract clicks, comments, shares, or time spent.

This design is powerful because it makes the network feel socially relevant. It is also narrow. It assumes that the best candidates are the things already circulating through the user’s existing relationships, and that the best ranking signal is engagement.

That is like designing a public library around two rules: only recommend books your friends have touched, and place the books most likely to provoke an immediate reaction at the entrance. The library might become very lively. It would not necessarily make its visitors wiser.

Engagement is not a useless metric. A person who saves an article, responds to a friend, or watches a tutorial to the end may be signaling genuine value. But engagement is an ambiguous measurement. Outrage can produce it. Compulsion can produce it. Familiarity can produce it. So can learning, inspiration, friendship, and practical help.

When one metric stands in for all these outcomes, the system begins optimizing the proxy rather than the purpose. The feed becomes very good at producing evidence that it is effective, even when its users are becoming less informed, less calm, or less able to act.

A ranking system does not merely reflect what people value. Over time, it teaches people what is available to value.

This is why changing the algorithm could matter more than changing the format. A photo, a short video, and a long essay are different containers. The deeper question is what kind of matching system decides who sees them, and why.

From the social graph to the possibility graph

The familiar social network is organized around the social graph: who knows whom, who follows whom, and who has interacted with whom. This graph is useful for maintaining relationships, but it is a poor map of everything a person might need.

Your closest friend may not know the answer to your question about starting a company in Brazil. Your family may not contain a specialist in pediatric sleep. Your followers may not have encountered the obscure historical essay that would change how you understand a current conflict. Limiting discovery to social proximity makes the network comfortable, but it also makes it intellectually provincial.

Information retrieval offers a more expansive mental model. In that model, a person can be treated as a query. The query includes interests, location, experience, constraints, goals, mood, and perhaps a temporary need. The things that might satisfy the query are documents, though documents need not be text. They can be videos, local services, research findings, conversations, software tools, courses, or people with relevant expertise.

This reframes recommendation. The system is no longer asking only, “What did your friends share?” It is asking, “What could meaningfully help this person now?”

That does not mean removing people. Human relationships are among the most valuable sources of trust, context, and care. It means refusing to treat the social graph as the entire universe of relevance. A good network might have several layers:

1. Relationship relevance: updates from people you care about.
2. Interest relevance: material connected to subjects you want to explore.
3. Life relevance: information related to your current circumstances, location, or goals.
4. Public value: discoveries and explanations that help people understand shared problems.
5. Serendipitous relevance: valuable material that does not fit your existing profile but expands it.

The last two layers are especially important. A system that only predicts what you already like will become an efficient machine for narrowing your world. A system that introduces carefully chosen difference can increase your capacity to notice opportunities.

This is where the idea of nonrival goods becomes central. Attention is scarce, but the objects attention can uncover are often not. A useful explanation can be copied. A scientific insight can travel. A successful institutional design can be adapted. A practical guide can help thousands of people without being consumed by the first reader.

The network’s job, then, is not simply to allocate scarce entertainment. It is to discover and distribute abundance that is currently hidden by poor matching.

The positive sum alternative to engagement

The dominant feed treats users as consumers competing for a limited supply of attention. Creators compete for distribution. Advertisers compete for access. The platform maximizes the amount of attention it can capture and resell.

A positive sum network would ask a different question: What is the user more able to do after encountering this?

That question changes the design of the product and the measurement of success. Instead of relying on engagement alone, a system could track signals such as:

• Did the person save the material and return to it later?
• Did they complete a task, learn a concept, or solve a problem?
• Did they discover someone or something they would not otherwise have found?
• Did the content improve the quality of a later decision?
• Did it lead to a durable relationship rather than a brief reaction?
• Did it leave the person more capable, rather than merely more stimulated?

These measures are harder to collect than clicks. They are also closer to the actual purpose of a useful information system.

Consider two hypothetical posts. The first provokes a furious argument that produces ten thousand comments. The second explains how a local community reduced antibiotic misuse, and is quietly saved by several hundred people who share it with teachers, doctors, and policymakers. The first looks superior under engagement metrics. The second may generate far more real value.

The difficulty is that positive effects often arrive slowly and indirectly. A person may read about a public health practice today, apply it months later, and influence an institution years after that. The path from information to social benefit is not a clean conversion funnel.

This is also why global public goods matter to the design of feeds. Climate stability, effective disease control, biodiversity, and financial stability cannot be protected by one person acting alone. They require cooperation across borders, institutions, and generations. Yet the information systems through which people encounter these issues often reward immediacy, identity conflict, and emotional escalation.

A network designed around public value would not need to become a government or a classroom. It would simply recognize that the most important things to distribute are not always the things most likely to trigger a reaction.

Finding the crack: where incumbents are structurally blind

Large platforms rarely lose because a competitor builds a slightly better version of the same product. They lose through neglected edges. An incumbent may be dominant on desktop but weak on mobile, excellent at text but indifferent to images, strong at existing relationships but poor at discovery.

The same logic applies to algorithmic competition. A new network does not have to beat an established platform at maximizing engagement across billions of users. It can begin with a neglected objective that the incumbent is structurally unable or unwilling to prioritize.

The opportunity may be a network for:

• Understanding complex topics without turning them into partisan spectacle.
• Connecting people with practical local knowledge.
• Helping researchers, builders, and teachers find collaborators.
• Matching citizens with credible explanations of shared problems.
• Creating spaces where users choose the kind of attention they want to spend.
• Rewarding material that improves decisions rather than merely extending sessions.

These are not just demographic niches. They are objective function niches. The product is differentiated by what it optimizes.

An incumbent whose revenue depends on maximizing time spent may find it difficult to build a network that encourages people to leave once they have accomplished their goal. A platform whose ranking system is tuned to emotional intensity may struggle to promote careful disagreement. A company organized around social identity may find it hard to recommend content that matters even when no friend has shared it.

That is the crack.

Early product strategy should therefore examine not only who is underserved, but also which valuable outcome the dominant system cannot measure. A small product can win with a modest audience if the members return because the system consistently helps them do something important.

This returns us to cohort thinking. Absolute user counts can be seductive, but they do not reveal whether a product is creating a durable behavior. A thousand users who return because each cohort finds increasing value may be more promising than fifty thousand users who arrive through novelty and disappear.

For a positive sum network, the crucial cohort metrics might include the growth of user capability, the diversity of useful connections, the rate of high quality discovery, and whether people become better at articulating what they need. Network effects should not merely make the network busier. They should make it more intelligent.

Building systems that enlarge the world

The practical lesson is not to abolish social media or pretend that engagement does not matter. It is to separate attention capture from value creation.

A better system could let users choose among modes: stay connected to friends, explore a subject, solve a problem, encounter unfamiliar perspectives, or receive updates on issues of public importance. It could make the objective visible rather than hiding it behind one opaque feed. It could also give users a way to correct the system by saying not only “show me less of this,” but “help me become the kind of person who understands more about that.”

Creators would need incentives that reward durable usefulness. A post might earn distribution because people cite it, build on it, translate it, teach from it, or use it to make a better decision. These signals would still be imperfect, but they would point toward a broader definition of success.

Institutions could participate as well. Universities, libraries, public agencies, and research communities produce vast amounts of valuable knowledge that rarely travels beyond specialized audiences. Better matching could connect that knowledge to people whose lives it affects, without requiring every person to become an expert or every institution to become an entertainer.

The deepest design principle is simple:

Build networks that increase the supply of useful possibilities, not merely the velocity of reactions.

Key Takeaways

• Treat the user as a changing query, not a fixed identity. Design for goals, needs, and curiosity, not only past clicks or existing friendships.
• Separate candidate generation from social proximity. Include material from expertise, public knowledge, local context, and constructive serendipity.
• Measure capability, not just engagement. Track saving, learning, completion, useful connections, and improved decisions alongside clicks and comments.
• Look for objective function gaps. The strongest opportunities often involve outcomes large platforms cannot prioritize because their business models reward attention capture.
• Evaluate cohorts by durable value. A healthy network makes each successive group more capable and better connected, not merely larger.

The future of social media may be less social in the narrow sense of showing us more people. That is not a retreat from human connection. It may be an expansion of what connection means.

A network can connect us to friends, but it can also connect a question to an answer, a problem to a method, a discovery to an institution, and one country’s successful experiment to another country’s urgent need. Those connections are often invisible because they do not fit neatly into the social graph.

The most important competition in digital media may therefore not be between platforms with different interfaces. It may be between different theories of what a person is for. One theory sees a user as a source of attention to be retained. Another sees a person as a set of open questions, capable of learning and contributing.

If we build around the second theory, the network stops being a machine that asks, “What will keep you here?” It becomes a public utility for asking a more valuable question: What could become possible for you, and for everyone connected to you, if the right idea found you in time?