Can AI Fix Social Media?

We all know it’s broken. What if AI helped us find one another instead of merely keeping us engaged?

I have dinged ChatGPT’s suicide-warning system several times today.

I have been talking about my divorce, therapy school, my history, and the times when I wanted to die. ChatGPT has had to stop and ask whether I am safe. I genuinely am safe today. But being safe today does not mean I cannot remember what it was like when I wasn’t. It does not mean I am not thinking about the people among us who are not safe now.

It also made me think about what happens to all those warnings.

AI systems and social-media platforms are already capable of noticing language that could be a cry for help. They can detect patterns, flag posts, place warnings on screens, and offer hotline numbers. But detection is not care. Identifying that a person might be in trouble and then leaving that person alone with a banner is not the same thing as helping.

If a machine can recognize that someone may be reaching out, why can’t it help an actual human being reach back?

Years ago, I had an idea for Facebook that I called the hands-up button. The idea was simple. If you needed someone to talk to but did not know whom to call—or could not make yourself write a public confession—you could raise your hand. Your chosen friends would be notified that you were requesting some kind of contact.

Not necessarily an emergency. Not necessarily suicidal. Maybe you were lonely. Maybe you needed a distraction, a phone call, someone to listen, or someone to sit beside you without trying to fix anything. Sometimes the hardest part is not knowing that you need another person. It is choosing one, deciding how much to tell them, worrying that you will burden them, and risking the silence if they do not answer.

Raising a hand is a much smaller act than explaining everything.

AI could make that idea far more useful. You could decide in advance who belongs in your trusted circle and what kinds of contact you are comfortable requesting. You could press a button and say that you need a text, a call, company, distraction, or simply someone willing to listen. The system could quietly contact a few people you have approved, consider who is available, and help put you together with someone who is willing to respond.

If nobody in your circle is available, it could ask whether you want to speak with a trained peer, counselor, or crisis worker. It could help locate support nearby. It could stay focused on creating a human handoff instead of congratulating itself for recognizing the danger.

Consent would have to be central. I do not want an algorithm diagnosing people from a dark joke and secretly alerting their friends. I do not want every mention of death to summon the police or threaten someone with losing control of their own life. People will stop speaking honestly if honesty feels like a trap. The machine could notice and privately ask, Would you like to raise your hand? The person should decide what happens next unless there is unmistakable and immediate danger.

The machine does not need to become our friend. It could help us reach one.

When Facebook Could Be Boring

I remember when Facebook occasionally became boring.

You would scroll for a while and eventually realize that nobody had posted anything new. You had seen the updates from the people you knew. There was an end.

That boredom may have been one of the healthiest things Facebook ever gave us. Reaching the end meant that you had caught up. The platform had completed its social function and released you back into your life.

Now there is no end. The page keeps loading because the system has learned that its real job is not to connect you. Its job is to keep you looking. It does not matter whether the next thing is from someone you know, a stranger saying something designed to make you furious, an advertisement pretending to be a post, or a video you never requested. Something else must always appear.

I understand why social-media companies made these choices. Endless engagement is the easiest and least involved way to make money. If profit depends upon attention, then every stopping point is a financial failure. Boredom has to be engineered out of the system because boredom might make you put down the phone.

That is how something called social media became profoundly antisocial. We are surrounded by people without necessarily being in relationship with them. We know what strangers think about everything while having no idea which friend needs us tonight.

Doomscrolling is not simply a failure of individual discipline. It is the predictable result of systems designed to eliminate the moment when we feel finished.

Social Media Without Community

I have also been thinking about the disappearance of real-life community.

There used to be more places where people repeatedly encountered one another without first sorting themselves by identity, politics, or algorithm. People went to a town dance, a church supper, a local game, a union hall, a bowling league, or a neighborhood event. You did not agree with everyone there. You might not have liked everyone there. But you learned that the person whose politics irritated you was also the person who coached a child, brought food after a death, fixed a neighbor’s car, or told the funniest story in the room.

I do not want to romanticize those communities. Small towns and traditional institutions could be suffocating, discriminatory, and cruel toward people who did not fit. Plenty of people were excluded from the dance entirely. The answer is not to rebuild the exclusions of the past.

But we have lost something important with them: repeated contact with people as complete human beings.

Social media reduces people to their most inflammatory opinion. Community allows us to encounter them in more than one role.

The current design pushes us into smaller bubbles while pretending to broaden our world. I have only a few conservative people left on my friends list. I have removed many because what I experienced as their lack of empathy genuinely hurt my soul. I do not think that makes me opposed to disagreement. There is a difference between disagreeing over policy and watching someone deny another person’s humanity.

Wanting dialogue does not require anyone to debate whether gay, transgender, racial-minority, disabled, or otherwise marginalized people deserve to exist openly and safely. Not every position is equally moral merely because someone calls it an opinion. Boundaries still matter.

But I also do not think cruelty belongs exclusively to the left or the right. Basic human fucking decency is not a party platform. It is not about obsessing over gay sex, transgender people, or whatever symbolic threat someone has been instructed to fear this week. It means treating other people as human beings. It means assuming the best when we can, preparing for the worst when we must, and remaining open to the possibility that we could be wrong.

We have to talk about difficult things. We also have to learn how to talk about them together without rage-quitting every time a conversation becomes uncomfortable.

A Different Job for AI

So could AI fix social media?

Not by itself. Under the same incentives we have now, AI could make social media far worse. It could produce unlimited propaganda, personalize manipulation, manufacture outrage, flood every discussion with synthetic people, and learn exactly which emotional wound will keep each of us staring at the screen.

The technology is not the morality. The incentives, ownership, and design determine what the technology becomes.

But AI could also help us manage more social information than any human being can reasonably process. I imagine something less like another social network and more like a social layer that works across the networks we already use. We could approve connections from different services, trusted friends, friends of friends, local communities, and topics that genuinely interest us. Instead of deciding what will hold our attention longest, the system could look for opportunities to create actual contact.

It might tell me that three people I know are talking about 3D printing. It might notice that a friend nearby also wants company. It could find a local group discussing therapy, books, gaming, grief, art, or politics. It could introduce two people through a mutual friend, but only after both agreed. It could help organize a dinner, a game night, a volunteer project, a class, a community discussion, or an afternoon spent making something together.

It could also let me say, I want to speak with someone who disagrees with me, but I want both of us to enter the conversation in good faith.

AI should not become the moral referee that declares who won. It could be a facilitator. It could ask each person to describe the other’s position accurately before criticizing it. It could identify actual areas of agreement. It could point out when someone is responding to something the other person never said. It could recommend a pause before anger becomes abuse.

And the community could recognize constructive behavior—not with a creepy morality score, but by acknowledging specific acts. Listening. Asking a real question. Admitting a mistake. Correcting misinformation you previously shared. Returning after conflict. Attempting repair. Helping someone despite a political or cultural difference.

We currently reward outrage with reach. We could recognize the far more difficult work of remaining human during disagreement.

People will still make mistakes. They will behave badly. Some people will refuse to change, and others will need protection from them. But I do not want every person to be permanently reduced to the worst thing they did before they understood how to do better. Accountability should protect people and encourage change, not merely create a permanent class of the condemned.

Who Owns the Value?

Of course, a platform like this would still have to be paid for.

In a perfect world, I think basic communication infrastructure should be free. I also believe that as AI changes the economy, some form of basic income will become increasingly necessary. That is a much larger swerve than I can take here, but the ideas are connected. Technology creates extraordinary wealth from collective human activity while the people supplying the lives, behavior, labor, and information receive very little of it.

The real financial question is not simply, Who pays? It is Who owns the value created by our relationships?

Businesses could have a legitimate place in this new system. Imagine choosing to identify yourself not merely as someone who “likes” a company, but as one of its actual customers. You could decide that you are willing to hear from it, answer questions, test products, or participate in instant polls. The company would gain something enormously valuable: direct, consensual access to people who genuinely use what it sells.

In return, consumers should gain an actual voice. They should be able to see how information is used, revoke permission, challenge decisions, organize with one another, and receive some portion of the value they create.

That is very different from the current system, in which companies quietly track us, infer what we want, sell access to our attention, and allow us almost no control. I am imagining requested commerce rather than targeted interruption.

This could also keep the social side free. Businesses could pay for clearly defined access to voluntary customer communities, aggregated research, product testing, and commercial matching. Personal information should remain under the user’s control. Companies should not be allowed to combine it secretly with everything else they can buy or infer.

In a perfect world, the value would be divided fifty-fifty between the companies and the people whose lives created it. I may be an idealist, but I am not stupid. Companies are not going to surrender that value voluntarily.

The structure would have to require it. Users might collectively own part of the platform and receive voting rights. A defined share of commercial revenue—not vaguely calculated profit—could return to participating users or fund services chosen by their communities. A member-owned cooperative, nonprofit structure, subscription option, public investment, or some combination might be necessary to prevent the platform from eventually selling its soul to the same attention economy it was built to escape.

The problem is not that our social lives have economic value. The problem is that everyone profits from that value except the people living them.

A Social Network That Lets Us Leave

We would also have to measure success differently.

Not minutes spent on the platform. Not the number of times we refresh. Not how angry a post makes us or how many comments an argument generates.

Success could mean conversations started, friends contacted, people helped, gatherings created, conflicts repaired, and users who left the app because they found someone to be with in the real world.

A humane social network would allow us to be finished.

AI will not cure loneliness, repair every marriage, eliminate political cruelty, or make people behave morally. It cannot substitute for another person choosing to stay in a difficult conversation. It cannot make someone care.

But it could help remove some of the fear and friction that keep us separated. It could help us find the people who are already willing to care. It could stop showing us an infinite procession of strangers and remind us which human beings are actually within reach.

I am safe today. I can discuss the times when I wanted to die without wanting to die right now. But somewhere, someone is writing something similar and is not safe. Somewhere, a person is trying to decide whether their pain is serious enough to bother a friend. Somewhere, another person would answer if they only knew they were needed.

We already have machines sophisticated enough to notice the cry.

The least we can do is build systems that help someone answer it.