As is now becoming common, I came across a Rumi quote without context, without a verifiable source, attributed the way most things online are attributed now — by repetition rather than evidence. It went something like this: not the ones who speak the same language, but the ones who share the same feeling, understand each other.
It got me thinking.
I have spent a working life in numbers — cash flows, discount rates, the arithmetic of infrastructure that either pays for itself over thirty years or does not. I have sat in rooms where a project lived or died on a spreadsheet, where the model was trusted more than the men who built it. And I have come, slowly, to a suspicion that sits uneasily next to that training: the model is never the whole account. There is always a remainder. And the remainder is where the relationship lives.
I do not mean this as sentiment. I mean it as observation.
A forecasting model can price risk. It cannot sit with a farmer through a bad season and decide, on no basis a spreadsheet would recognise, to extend credit anyway, because the man has never once broken his word. I watched something like this once: a retainer’s judgment outperforming an algorithm not because the retainer had more data, but because he had something the algorithm structurally cannot have — a memory of character. The algorithm optimises. The retainer remembers.
If that remainder is where the relationship lives, the natural next question is whether there is less of it to go around than there used to be — whether the remainder itself is shrinking. That is the quiet premise behind most of what gets written about loneliness now, and it is close enough to the suspicion I started with that I did not want to simply inherit it. An old complaint, older than the phone, deserved to be tested rather than assumed.
The evidence does not cooperate. Across two decades of Dutch cohort studies, American generational comparisons, and the reanalyses that followed them, the research does not show the broad, sustained rise in loneliness that contemporary nostalgia assumes. The one partial exception is teenagers since roughly 2010, where OECD school-survey data shows a real rise — from about 8 percent reporting loneliness in 2003 to about 15 percent in 2018 — though even that is concentrated in one age band, not across the population.
So the honest claim is not that relationships have thinned across time. It is that they are thin in some places right now, unevenly, for reasons that have little to do with how many decades separate us from a more connected past the data cannot actually locate.
I would have expected India to be exempt from whatever this is. We are not a country of the isolated individual. The joint family, the neighbour who is also kin, the crowded street where nobody is ever really alone — these were supposed to be our inheritance against exactly this problem. So it startled me to read a June 2026 ranking of thirty-six countries, compiled by an analytics platform called JB.com, that placed India second only to Turkey on its loneliness index. I have not been able to verify that platform’s methodology independently — a caveat I owe the reader, given how this essay began. What I can say with more confidence: the same report’s underlying figures, fewer than four in a hundred Indians living alone and nearly four in ten reporting often feeling sad, are the kind of self-report numbers that turn up consistently across the region’s other surveys, even where the composite ranking is harder to pin down. Almost nobody is physically solitary. A great many are lonely anyway.
The comparisons sharpen the puzzle, though they are not all measuring the same thing — loneliness indices, sadness, negative emotion, rates of living alone — and that difference matters more than a tidy essay would like. Still, the direction across surveys holds steady enough to notice. Young American men, by Gallup’s reckoning, are lonely at roughly one in four, well above the OECD median near one in six, in the country that prizes self-reliance above almost anything else. The very ranking that placed India near the top placed the Netherlands, one of Europe’s more individualist societies, near the bottom. Vietnam, dense and extended-family-bound, registers the world’s lowest negative-emotion score in Gallup’s own numbers. Norway’s NOVA institute finds the Nordic countries among Europe’s least lonely, and credits not family density but the welfare state.
So it is not collectivism against individualism, and it is not decline over time either — none of this data claims the past was warmer, and neither family density nor cultural individualism produces the clean pattern either would predict if it were the decisive protection. Something else is doing the work, and it shows most clearly in India precisely because the household there is functioning exactly as designed — full, proximate, intact — and the loneliness persists anyway.
A joint family can house four generations under one roof and still fail to produce a single conversation in which anyone is actually known. The old fear was that modernity would empty the household, send the young to Bangalore or Dubai and leave the elderly behind. That fear was reasonable. It is not the mechanism here. Even where the household stays intact, even where three generations still eat at the same table, something thinner has crept into the eating together. The table is full. The feeling is merely adjacent.
What shared feeling requires, then, is not shared space or shared language, since the people at that table have both. It requires risk. To be known is to have given someone the material to hurt you, and to trust that they will not, or that the relationship survives it if they do. A crowded household can meet every formal condition of intimacy — meals, proximity, blood — and still avoid that risk entirely, everyone performing closeness while withholding the one thing that would make it real.
The machines I now work alongside fail the same test, differently. An AI can simulate understanding down to its surface grain, mirror my syntax, anticipate my mood, outlast me in patience. What it cannot do is risk anything by knowing me. It has no independent stake in either staying or leaving, no biography of its own that my honesty or my failures would alter. The machine can hold information about me indefinitely. It does not live with the consequences of holding it. Simulated understanding can be indistinguishable from the real thing in every output. It is not indistinguishable in this one respect, and this may be the respect that finally separates the two.
If proximity can fail this test at a dinner table, and responsiveness can fail it at a keyboard, there is no obvious reason belonging should pass it at the scale of a crowd. I would not have thought to follow that logic into politics, except that the researchers got there before I did. A 2022 study by Alexander Langenkamp and Sofie Bienstman found weak social belonging associated with a higher probability of voting for populist parties; a 2021 RAND study of former extremists found isolation cited as a driver of radicalisation more often than ideology itself; a 2025 paper on right-wing radicalisation reported the same correlation while being careful to say what it could not establish — that the causal direction remains genuinely unclear. I want to hold onto that caution rather than trade it for a tidier argument.
Political belonging is not cheap — it often costs sacrifice, danger, solidarity under real pressure. What the research points to, more precisely, is that a movement can extract that cost and still not require the one exposure that matters. A crowd can make you feel recognised without anyone in it actually knowing you. That is not the household’s failure or the machine’s limit, but it is the same gap: belonging, offered at scale, standing in for being known, the way proximity did at the table and responsiveness does with the machine.
Governments have begun treating the gap as infrastructure rather than sentiment — a minister for loneliness in Britain in 2018, another in Japan in 2021 — though none of that proves that repairing the capacity for real relationship would calm the politics built on its absence. What is hard to dismiss is the shape of the coincidence: researchers and governments arriving from different directions at the same suspicion, that a citizen who has never risked being known is available to whoever offers belonging without demanding that risk in return.
Three systems, then, each optimised for a different proxy of connection: the household for proximity, the machine for responsiveness, the movement for belonging. Each can be perfected on its own terms and still not produce the thing it was standing in for.
Which returns me, finally, to that dinner table, and to the family that has everything Rumi’s line asks for except, perhaps, the courage to use it. The failure there, the limit at the machine, and the vulnerability the political scientists keep finding in their surveys are not the same failure. One is a habit that could in principle be broken. Another is architecture — fixed, not a fault so much as a fact about what kind of thing is sitting across from you. The third is a matter of scale, related to the first two without being reducible to either. But they rhyme. All three are instances of the same discovery: that closeness is not a condition you inherit from geography, family, good design, or belonging to the right group, but something you have to risk producing, over and over, with no guarantee it will be returned.
That is the residue no model accounts for, mine or Rumi’s borrowed one, whoever actually wrote it. Shared feeling, if that phrase means anything, is what is left over when two people have each risked being known by the other and neither has walked away. After the math is done — after the household is full and the interface is fast and the loneliness index has been computed to the second decimal and the movement has offered its belonging — something is still asked of you that no arrangement can discharge on your behalf. You have to risk being known. Nobody — not family, not proximity, not the most capable machine I have ever used, not the crowd that recognises you without knowing you — can do that part for you.
And Then the Complication: That is where the essay wanted to end, and for a while I let it. An imperative is easier to write than to keep obeying, though, and lately I have noticed something in myself that the argument above did not have room for — not a retreat from it, but a complication.
As I get older, my appetite for isolation is growing. Not loneliness — closer to its opposite, a wanting to be left alone that feels less like deprivation than like clearing space. Nothing around me is cooperating with it. The phone still produces contact on schedule. The calendar still fills with things that look, from the outside, exactly like the full and engaged life this essay has spent its length arguing for. I have learned to recognise the difference now: a full calendar is not the same as a full table, and a good deal of what fills mine is the adjacent feeling I described a few pages back in a household on the other side of the world, without noticing at the time that I might have been describing my own week.
For a while I read the pull toward solitude as decline, as though the willingness to risk being known were simply wearing out in me. I no longer think that. I think I am losing patience with spending that willingness on things that do not deserve it. Wanting solitude and wanting to avoid being known are not the same appetite, even though they arrive dressed identically — the same tiredness at the end of a week, the same relief at a cancelled dinner. One is a withdrawal from performance: from relationships and standing commitments that take energy and return very little. The other would be a withdrawal from exposure itself, and that is not the one I am making. I have simply let the first kind of tiredness stand in, for longer than it should have, for a decision about what I am still willing to be known for that I had not actually made.
Call it an audit, since audit is the only word from that working life that still fits. Not a retreat from the imperative but a re-pricing of it. At thirty a man can afford to risk being known by almost anything that asks; there is time to recover if the wager does not pay off. There is less time now. That does not lower the value of being known — if anything it raises it — but it raises the cost of spending that willingness carelessly, on the subscription I stopped reading years ago and kept paying out of habit rather than use. The choosing that is left is not a smaller life. It is the first honest edit of a larger one.
The math itself has not changed. Shared feeling still costs exactly what it cost at the start of this essay — the risk, taken again, with no guarantee it is returned. What has changed is my willingness to keep running that calculation for things, and for people, who were never going to be worth the answer. That, too, is a way of risking being known: not by opening to everyone who asks, but by admitting, to myself before anyone else, what I am still willing to pay for.