Shared Reality Is Cracking. Here's What Rebuilds It

The Sudden Rise of "Shared Reality"

I've noticed the phrase showing up in places it never used to. Newsletters, panel discussions, casual arguments with family. A few years ago people said "we disagree." Now they say "we don't even share a reality anymore," as if the disagreement itself were the smaller problem and the missing common ground were the larger one.

That shift in language is worth taking seriously, because it points at something more specific than polarization. Polarization is about distance between positions. Shared reality is about whether two people are even standing on the same ground before the distance gets measured. You can disagree sharply with someone and still share a reality with them, if you agree on what happened, what the words mean, and what counts as evidence. What's newly alarming to a lot of people isn't the disagreement. It's the sense that the floor itself is different for each side.

In my view, the current wave of attention is being driven by three things landing at once: AI systems that can generate a convincing account of almost anything, feeds that quietly sort people into different factual universes, and a political moment where competing camps increasingly can't agree on a shared set of facts to argue about. None of these are new problems by themselves. What's new is how fast they're compounding.

The Actual Meaning Behind the Phrase

The phrase has a real intellectual home, and it's older and more precise than its current popular usage suggests. The social psychologist E. Tory Higgins built an entire research program around it, culminating in his 2019 book Shared Reality: What Makes Us Strong and Happy (Oxford University Press). Higgins's core claim is that humans don't just perceive the world privately and then compare notes. We actively seek to establish that our experience of something is shared with another person, and that act of establishing agreement is itself what makes an experience feel real and significant. A sunset isn't fully real to you, in his framework, until someone else confirms they saw it too.

That's a useful definition because it reframes the current anxiety. The problem isn't that people hold different opinions. It's that the everyday mechanism for confirming a shared experience, the small back-and-forth of "did you see that" and "yes, I saw it too," is breaking down at scale. When that mechanism breaks, opinions don't just diverge. The underlying sense that anything is confirmed at all starts to erode.

This isn't a new worry dressed in new language. Walter Lippmann was writing about it in 1922, in Public Opinion, when he described how people act not on the world itself but on "the pictures in our heads" of the world, pictures assembled from whatever partial information reaches them. Lippmann's insight was that this gap between the world and the picture of it is permanent. What's changed is the machinery that assembles the pictures, and how much less overlap there now is between one person's picture and another's.

Why Shared Reality Is Collapsing Now

I think the honest answer has to hold two things at once: the platforms that sort us, and the humans doing the sorting.

Cass Sunstein made the platform-side argument early, in his 2001 book Republic.com, when he warned that a world where people could filter out disagreeable information would produce group polarization rather than deliberation. He was writing about the early internet's promise of infinite customization, before algorithmic recommendation existed in its current form. What he predicted as a possibility has become closer to a default setting: feeds that learn what keeps you engaged tend to learn what confirms you, because confirmation is comfortable and comfort is sticky.

But the philosopher C. Thi Nguyen made a distinction in 2020 that I think gets lost in most popular discussion of this problem, in a paper called "Echo Chambers and Epistemic Bubbles," published in the journal Episteme (vol. 17, no. 2, 2020, pp. 141–161). Nguyen separates two things that look similar from outside but require different fixes. An epistemic bubble is a structure where you simply haven't been exposed to opposing views, an information gap. Pop the bubble by showing someone the other side's argument, and the problem can genuinely resolve. An echo chamber is a structure where you have been actively taught to distrust outside sources, so showing someone the other side's argument doesn't help. It confirms the suspicion that the other side is dishonest, and the chamber gets stronger.

That distinction matters enormously right now, because a lot of well-meaning efforts to "restore shared reality" are epistemic-bubble interventions, more exposure, more fact-checking, more cross-cutting content, aimed at what is often actually an echo-chamber problem. You cannot pop an echo chamber by feeding it more facts. The facts are exactly what it's built to reject.

Layer generative AI on top of this and the picture gets harder still. It isn't only that AI can produce convincing fabrications, though it can. It's that AI has made the cost of producing a plausible, tailored account of events approach zero, at the same moment the cost of verifying any given account has stayed roughly where it was. The asymmetry between how cheap it is to generate a claim and how expensive it is to check one is, I think, the sharpest edge of the current moment.

Three ways people try to rebuild it

Not every approach to restoring shared reality works the same way, and I think the differences are worth laying out plainly rather than treating "just talk more" as a single strategy.

I put debate and fact-checking in that table not because they're worthless, but because they solve for a different problem than the one most people are actually facing when they say "we don't share a reality anymore." Debate assumes the two parties already agree on the facts and are contesting the conclusion. Fact-checking assumes one party is simply missing information. Reflective listening starts from a different premise: that before you can even locate the disagreement, you have to establish, out loud, that each person has correctly understood what the other believes and why. That confirmed understanding is Higgins's shared reality in miniature, rebuilt one conversation at a time instead of assumed as a background condition.

What history tells us about this fight

Hannah Arendt wrote an essay in 1967 called "Truth and Politics," originally published in The New Yorker on February 25, 1967, and later collected in Between Past and Future, that I think is the sharpest thing ever written on this exact anxiety. Her argument was that factual truth, unlike philosophical truth, is always vulnerable to power, because factual truth depends entirely on the testimony of witnesses and could always have been otherwise.

A politician cannot argue you out of two plus two equaling four, but a politician absolutely can get enough people to doubt what actually happened in a room, on a given day, because the alternative to that fact was always logically possible, just not actual. Arendt thought this made factual truth far less self-evident than mathematical truth, and far more exposed to organized lying.

What I find sobering about her essay, reading it now, is that she wasn't describing social media. She was describing a much older vulnerability in human knowledge that new tools have simply made faster and cheaper to exploit. Shared reality was never guaranteed by nature. It was always a fragile, ongoing achievement, propped up by institutions, witnesses, and a baseline willingness to trust testimony. What's changed is the scale at which that achievement can now be undermined, and the speed at which it happens.

What actually works, in my experience

I don't think the answer is a platform fix or a policy fix, though both might help at the margins. I think the answer is closer to what Higgins was describing: shared reality gets built person to person, through the deliberate, unglamorous work of confirming that you've understood someone before you respond to them.

That's a slower process than debate and a less satisfying one than fact-checking, because it doesn't produce a winner. What it produces is something more foundational: two people who now stand on the same piece of ground, even if they still disagree about what to build on it. I've come to think that's the actual prerequisite for any disagreement worth having. You can't productively argue about what to do next if you can't first agree on what just happened.

This is why structure matters more than good intentions in these conversations. Two people who trust each other and wing it will usually default to their most habitual pattern under stress, which for most of us is defending our own account rather than confirming the other person's. A simple, repeatable structure, something as basic as restating someone's position back to them before responding to it, does the work that goodwill alone tends not to.

Here's what that looks like in practice. If someone says "you never take the data seriously," the instinct is to defend yourself or produce the data. Reflective listening asks you to do something slower first: "Before I respond, let me check I have this right — you've presented data before and felt it got waved off rather than actually engaged with. Is that a fair account?" Only after they confirm or correct that do you respond to the substance. The restatement isn't a stalling tactic. It's the mechanism: you earn the right to be heard on the substance by first proving you received what was actually said.

It's a small ritual, but rituals are how shared reality gets rebuilt in practice rather than just in theory. I've written more about how that kind of structure keeps multiple perspectives in view at once without collapsing into one side's account in an earlier piece on the parallax effect, and about why the written rules of a conversation tend to outperform whatever norms everyone assumes but never states in this piece on written dialogue rules.

None of this fixes the platform incentives or the economics of generated content. I don't think any individual practice can. But it does mean the erosion of shared reality isn't purely a structural problem happening to us. It's also, still, something built or unbuilt in the specific conversations we choose to have carefully instead of quickly.

Frequently Asked Questions

Is "shared reality" just another word for agreement? No. Shared reality is about agreeing on what happened and what something means before any conclusion is drawn from it. You can share a reality with someone and still disagree sharply about what to do about it. What breaks down when people say they no longer share a reality is the earlier, more basic layer: the confirmed sense that both people experienced or understood the same event.

Is AI making the shared reality problem worse? Not by inventing the problem, but by changing its economics — see above for why the widening gap between generating and verifying a claim matters. The practical upshot: any single fact-check now buys less confidence than it used to, because the next fabrication is already cheaper to produce than the last one was to debunk.

What's the difference between an echo chamber and an epistemic bubble? See the distinction above, from C. Thi Nguyen's 2020 Episteme paper. The practical test: ask whether more exposure to the other side's argument would actually change anything. If the honest answer is "that would help," you're looking at a bubble. If the honest answer is "that would just prove they're all in on it," you're looking at a chamber — and more facts alone won't fix that.

Can two people rebuild shared reality without agreeing on politics? Yes, and I'd argue that's the more common and more useful case. Rebuilding shared reality means confirming that each person has accurately understood the other's position and experience. It doesn't require either person to change their conclusion. In fact, disagreement tends to become more honest and more productive once both sides trust that they've actually been heard correctly.

Where does shared reality actually get rebuilt, if not through policy or platform design? In individual conversations, through structure rather than good intentions alone. E. Tory Higgins's research suggests shared reality is established through the active, mutual confirmation that an experience is understood the same way by both parties. That's a practice, not a platform feature, and it can be built into how a specific conversation is run.

Last updated: 2026-09-12