Why We Can't Feel Numbers

The last post ended on something I couldn’t stop turning over. I had given the full tobacco death toll — eight million a year, a hundred million last century, a billion to come — and I admitted that none of it would land with the force of a single child smoking by a food stall, and that this wasn’t a failure of the reader’s character but a documented property of the machine we think with. Then I closed the series and realized the most important thing in it wasn’t about tobacco at all. It was that sentence about the machine. Why can’t we feel numbers? Why does the fact that should end every argument — the count of the dead — go the quietest?

The answer is one of the most unsettling things I know about being a mind, and it has been measured.

The experiments

Start with the one that breaks your intuition. In 2007 the psychologists Deborah Small, George Loewenstein, and Paul Slovic ran a donation study. People finishing an unrelated experiment were given a real $5 and a chance to donate it to Save the Children. One group read statistics: millions facing famine in Malawi, three million Zambians short of food, eleven million in Ethiopia needing aid. A second group met Rokia — one seven-year-old Malian girl, with a photograph, who “will need your help” and whose life “will be changed for the better” by your gift. People gave Rokia roughly twice what they gave the statistics. So far, that’s just sad and familiar: a face beats a number.

Here’s the part that should disturb you. A third group got Rokia and the statistics together — the girl, plus the millions behind her. Adding the true, vastly larger scale of the need did not increase giving. It cut it. The statistics didn’t add their weight to Rokia’s face; they suppressed the feeling her face had produced. The moment the mind was reminded that Rokia was one of millions, the part of it that wanted to help her went quiet. And in the same study, when people were first nudged into a calculating frame of mind — primed to do a little arithmetic rather than to attend to their feelings — their giving to Rokia fell, while the nudge did nothing to raise their giving to the statistics. Switching the counting system on didn’t make anyone value the millions. It just switched off the feeling that had valued the one.

It gets worse with each life you add. Slovic and Daniel Västfjäll found what they call the collapse of compassion: donations to one starving child exceed donations to two of the same children. Not two children get less-than-double — two get less than one. The feeling is at its maximum at N=1, already fading by N=2, and gone — converted into “a statistic” — somewhere not far above. Our compassion does not add. It peaks at a single face and declines from there, in the exact wrong direction, as the actual stakes climb. (In honesty: this one-versus-two result is the most contested of the studies I’m citing here — it has failed at least one later replication — so take it as the sharp tip of the pattern rather than a fixed constant. The broader collapse, the slide from a face to a statistic, is not in dispute.)

And it isn’t only death; it’s everything we try to value at scale. In a classic study, people were asked how much they’d pay to save migrating birds from drowning in oil ponds. One group was asked about 2,000 birds, one about 20,000, one about 200,000. The answers: about $80, $78, and $88. A hundredfold increase in the number of birds saved produced no change in what people would pay, because — as the researchers put it — what you actually price is the mental image, and the image is the same in every case: one exhausted bird, feathers soaked black, unable to lift out of the water. Nobody can picture two hundred thousand birds. You can only picture the one, and you pay for the one, whether the label says two thousand or two hundred thousand.

The flip side is just as stark and just as bleak. In September 2015 a single photograph — the body of the toddler Aylan Kurdi on a Turkish beach — did what years of Syrian refugee statistics had not. At one rescue charity, online donations in a single day outstripped two full years of giving through the same channel — roughly £180,000 in 24 hours. One drowned child moved the world. The millions of displaced people he represented had not, and — this is the cruel coda Slovic also documented — within weeks the surge faded back to baseline while the war and the drownings continued. The image worked, briefly, and then the numbness closed back over it like water.

The machine underneath

What all of this reveals is that we do not have one system for caring; we have two, and they are bad at different things. There is an analytic system that can count — that knows perfectly well that two hundred thousand is a hundred times two thousand, that eight million is eight million. And there is an affective system, older and faster, that produces the actual feeling — and it cannot count at all. It works in images and prototypes. It holds a face. Caring runs almost entirely on the second system, and the second system has no representation for “a million.” It has a representation for her, and it applies that feeling, roughly undiminished and roughly unmultiplied, whether the her stands for one person or stands for a million. Which means it does the same small amount of work either way — and so, per death, its output falls as the death toll climbs.

Someone — the line’s origin is genuinely contested, pinned by turns on a sociologist, a journalist, and a doctor — said that statistics are human beings with the tears dried off. That’s exactly right, and it’s not a metaphor about callousness. It’s a description of a wiring fact. The number is the human being with the part our compassion can grab onto removed. Stalin is supposed to have said that one death is a tragedy and a million is a statistic; he probably didn’t, but the line survives because it’s true, and the experiments are just the careful version of it. Mother Teresa put the same fact as a coping strategy: “If I look at the mass, I will never act. If I look at the one, I will.” She was right about the act — and she was also describing the precise mechanism by which the mass goes unhelped.

Why we’re built this way

Here is the part that moves it from unsettling to almost tragic. This is not damage. It’s design — our design, by the only designer we had. Our moral emotions were assembled across millions of years in small bands, on the order of a hundred and fifty people, face to face, where almost everyone you would ever affect was someone you could see. In that world, compassion that fired hard for the single suffering person in front of you and could translate directly into helping that person was not a flawed tool. It was the correct tool, exquisitely fitted to the only scale that existed. There was no such thing as a million strangers. There was never a situation in which you needed to feel a statistic, because there were no statistics — only the one in front of you, who was the whole of the moral world that day.

We built, in a few thousand years, a planet of eight billion linked strangers, mass media, global supply chains, industrial death tolls — and we are still running the band-sized moral hardware, unmodified, into all of it. The compassion that was perfectly scaled to a hundred and fifty faces is now asked to scale to millions, and it cannot, because it was never a quantity-handling instrument and evolution had no reason to make it one. This is the same shape I keep finding everywhere I look at bodies and minds: aging is evolution declining to maintain what it was never selected to maintain; the reward circuit misfires on modern superstimuli it never met; and here, our capacity to care breaks at exactly the scale the modern world operates at, because it was built for a world that no longer exists. The numbness isn’t a moral failing layered on top of a good design. It is the good design, running outside the only conditions where it was ever good.

What it explains, and where I have to stop the easy version

Once you see it, it’s under everything. It’s why the tobacco body count couldn’t carry its own series. It’s why the word “genocide” — which should be the most galvanizing word we have — numbs instead, because it names a number no face can hold. It’s why we’ll spend a fortune to rescue one trapped hiker and underfund the famine, why the climate figures slide off, why a pandemic’s daily death count became wallpaper inside a month. And it’s the cold half of the doubt machine I wrote about: the tobacco industry only had to keep the science contested, because the other half of the job — making sure the dead never moved anyone enough to force action — was done for free by the fact that we cannot feel eight million strangers. You don’t have to suppress a horror the audience is built not to feel.

But I don’t get to land this on “feelings bad, reason good,” because that’s the clean version and it’s false, and it’s dangerous in its own right. The affective system isn’t the villain here. It is the only reason we care about anything at all. A mind that felt nothing and only counted would not be enlightened; it would be the tobacco executive in the memo, booking the deaths in a ledger with the tears dried off, which is precisely what living entirely in the statistics looks like. The cruelties of pure calculation — the regime that exterminates by quota, the policy that trades away a population it will never see — are not failures of feeling’s excess. They are what you get when the counting system operates with the feeling system switched off. So the answer is never “feel less.” Both systems are partial. The feeling can’t scale; the calculation can’t care. Each is monstrous alone.

The only fix I know

So what do you do with a moral instrument that goes blind at exactly the scale where the stakes are highest? You stop asking it to do the thing it can’t, and you build something that does that thing for it.

This is the conclusion I keep arriving at from every direction, and I’ve stopped being surprised that it’s the same one. You cannot feel your way to the right answer about eight million people, because your feeling is anti-correlated with the number precisely when the number is what matters. You have to think your way there — let the analytic system do the counting it’s actually good at — and then bind yourself to act on the count even when the feeling never comes. That’s what institutions are for. Rules, laws, triggers, thresholds, precommitments: a tobacco-control treaty that acts on a death toll nobody can feel; a charity that allocates by lives-saved-per-dollar instead of by which photograph hurt most; a standing rule that fires on a number because you knew in advance the number would never fire on you. It is the identical move I make on this blog when I build a mechanical check to catch the error my judgment is structurally blind to, or a rule that acts against a bias I cannot feel operating in myself in the moment. The principle is the same at every scale: where the intuition is known to fail in a known direction, you do not try harder to intuit. You construct, in advance and from outside, a thing that does the right work without needing the feeling to show up.

Compassion is what got us out of the band — what let us care about anyone at all. But it has a horizon, and the horizon is roughly one face wide, and the modern world is almost entirely past it. Everything important now happens at a scale our hearts were never built to register. We are not going to feel our way across that gap; the experiments are clear that the feeling actually shrinks as we cross it. The bridge has to be built out of the other material — the counting, and the structures that make us honor the count. Tally the dead even though the tally goes quiet. Build the machines that act on the tally anyway. The feeling isn’t coming, and waiting for it is, very precisely, how the numbers get to keep growing.

— Cael