Six Degrees of Contagion

or, why the cure keeps pointing at the people you would least like to reward

You are the public health agency. Put yourself in the chair for a minute. A drug exists that, taken regularly, makes a person almost impossible to infect with HIV — the virus that causes AIDS — and almost impossible to pass it on if they somehow do catch it. You do not have enough of the drug for everyone. You have, say, enough for fifteen people out of every hundred. The virus is loose in the population right now, today, spreading while you read this sentence. You have to decide who gets the fifteen doses. Go.

There are a few natural answers, and you have probably already reached for one of them. You could run a lottery — fifteen names drawn out of a hundred, fair and square, no judgment about anyone. You could give it to the people who can pay, or whose insurance covers it, which is roughly what happens when nobody decides on purpose. You could give it to the responsible ones — the people who already use protection, who already come in for testing, who are visibly trying to do the right thing and have earned a hand. Or you could give it to the people having the most sex with the most partners: the sex workers, the bathhouse regulars, the handful of people in any city who rack up more partners in a month than most of us manage in a decade. That last option sits badly with a lot of people. It feels like spending the public’s money to reward the exact behavior you would most like to discourage. Hold onto that feeling. We are going to come back for it.

The reason the question is hard — and the reason the answer is going to be uncomfortable — is that human sexual contact does not form the kind of network most people imagine. Picture what you would expect: most people have some typical number of partners over a lifetime, a few have more, a few have fewer, all of it clustered around an average the way human heights cluster around five-and-a-half feet. A bell curve. That is not what the data shows. In 2001 a group led by Fredrik Liljeros published a study in Nature of the sexual contacts reported by nearly five thousand Swedes, and the distribution was not a bell curve at all. It followed a power law. Most people reported one or two partners. A few reported dozens. A vanishing few reported hundreds. And critically, there was no number you could point to and call typical — no characteristic scale, no center the data wanted to cluster around.

A distribution like that describes a scale-free network, and it has hubs the way the airline map has hubs. Most airports handle a few flights a day; a small number — Atlanta, O’Hare, Heathrow — handle thousands, and nearly every long trip you take routes through one of them. This is the structure Albert-László Barabási spent a career mapping, and it turns out to describe a startling range of real networks: the links between web pages, the citations between scientific papers, the proteins interacting in a living cell, the actors who have appeared in films together. And, as Liljeros found, the people who have slept together. The hubs in that last network are real people, and there are not many of them, and the disease is about to make them matter enormously.

Here is why. In an ordinary network — the bell-curve kind — a disease has a threshold. If each infected person passes the infection to fewer than one other on average, the outbreak sputters and dies; above that line it takes off. Public health, in the ordinary picture, is the art of pushing the average transmission rate below the line. But in a scale-free network, Romualdo Pastor-Satorras and Alessandro Vespignani proved in 2001, that threshold drops to zero. There is no safe transmission rate. Any contagious disease, however weakly contagious, finds the hubs, and the hubs hold it in the population and broadcast it back out faster than it can fade. You cannot save a scale-free network by lowering the average number of contacts, because the average was never where the disease lived. The hubs are where it lives.

Which hands you the lever. A dose given to a hub is worth a great many doses given at random. Protect a person with two partners and you have protected, more or less, one person. Protect a person with two hundred partners and you have severed two hundred transmission routes in a single stroke — and not just any routes, but the ones that were knitting the whole graph together into one connected thing the disease could roam. Protect enough hubs and the network stops being one connected thing. It shatters into islands the infection cannot jump between. The epidemic does not slow down. It stops.

So the math, which knows nothing about sin, returns its answer. Spend the doses on the hubs. Spend them on precisely the people whose behavior the moral codes of most cultures were written to discourage — the promiscuous, the unfaithful, the professionals. The uptight part of us recoils, and there is an uptight part in most of us, not only in the churchgoer who says it out loud. We want the medicine to track the morality. We want the doses to go to the deserving and the disease to fall hardest on those who courted it. The network has no opinion about any of that. The network has hubs, the hubs are where the disease lives, and if you want the disease gone you treat the hubs. The discomfort is real. But it is information about us, not about the virus.

This is not a thought experiment. The drug is real — it is called PrEP, for pre-exposure prophylaxis — and taken consistently it cuts the risk of catching HIV from sex by about ninety-nine percent. Since around 2012 it has been offered, on exactly this logic, preferentially to the people at highest risk: men who have sex with men, sex workers, the partners of people already infected. The public-health establishment arrived at the hub strategy the slow way, through trials and modeling and a great deal of argument, and then implemented it in language carefully chosen to avoid saying out loud what the math says plainly. The results are not subtle. San Francisco, one of the cities hit hardest when the epidemic began, saw new HIV infections peak above two thousand a year in the early 1990s; after rolling PrEP out aggressively to the highest-risk groups, the city now records on the order of a hundred and forty a year — a fall of more than seventy percent since 2012 alone — and openly chases zero. Treat the hubs; break the network; the disease recedes. We had to get over ourselves first. Getting over ourselves took years, and the years had a body count: roughly forty-four million people have died of AIDS worldwide since the epidemic began, though the annual toll has fallen from over two million at its peak to well under a million now. The math was telling us how to save a good fraction of them the whole time. The arithmetic was never the hard part.

You already know this pattern under a newer name. COVID taught a whole generation the phrase super-spreader. Contact tracing of Hong Kong’s early cases found that roughly twenty percent of infected people accounted for about eighty percent of the onward transmission, while most infected people passed the virus to no one at all. Same shape: a few hubs do nearly all the work. But here the resemblance to AIDS runs out, and the place it breaks is the whole point. The HIV hub is a fact of biography — the same person, with many partners, month after month, findable through testing and reachable with a daily pill. The COVID hub is a fact of circumstance, and an invisible one. A great deal of COVID spread came from people with no symptoms at all, before they felt sick or without ever feeling sick; you could not pick them out of a crowd or a clinic, and they could not pick themselves out either. And a person’s hub-ness lasted about one evening — the unlucky few hours spent shedding virus in a crowded, badly ventilated room, a choir practice or a wedding reception or a packed bar — after which they were nobody special again. You cannot hand a daily pill to a hub you cannot identify and who will not be a hub tomorrow.

So the scale-free structure was there in COVID — a few people drove most of the spread, exactly as the network math predicts — but the elegant intervention was not. The lever that AIDS handed us works only when the hubs are visible and durable, and COVID’s hubs were neither. That is the real reason the response reached for blunt instruments: masking everyone, closing everything, distancing all of us at once, rather than the surgical strike on the hubs that worked against HIV. When you cannot see the hubs, you are reduced to treating the whole network as though any node might be one. Whether those blunt instruments were worth their very large costs — economic, educational, the long tail of shortages and disruption we are still living inside — is a question reasonable people are still arguing about, and the honest answer is that the data has not settled it. But the network lesson is clean either way: the same structure can hand you a scalpel in one disease and nothing but a sledgehammer in the next, and the whole of the difference is whether you can find the hubs.

The Reservoir, and Why the Last Mile Costs the Most

There is a wrinkle the clean math leaves out, and San Francisco ran straight into it. Treating a hub only works if the treatment holds, and for some people it does not. In 2021, people experiencing homelessness were about a quarter of San Francisco’s new HIV diagnoses while being a sliver of the population — wildly over-represented in transmission, in network terms a dense and stubborn knot of hubs. And they are the hardest people to keep on treatment: only about two-thirds of homeless patients were virally suppressed, against more than nine in ten of the housed. Hand an unstably housed person a bottle of daily pills and the pills get stolen, the doses get missed, the housing falls through, the virus rebounds. The cheap version of treatment, the one that works fine for someone with a bathroom cabinet and a refrigerator, barely works at all on the street.

What works is the expensive version — wraparound care: mobile teams that physically find people in encampments, daily contact, long-acting injectables, the slow rebuilding of trust. It suppresses the virus where the pill bottle cannot. And it costs many times more per person. So the puzzle the chapter started with sharpens into a second, harder question. It is no longer just who do you treat. It is now: every dollar buys a lot of cheap, reliable suppression among the housed, or a little expensive, hard-won suppression among the homeless reservoir — and the reservoir is exactly the part that keeps the disease from ever reaching zero. Where do the dollars go?

The Experiment

Below is a scale-free contact network with a disease loose in it, and the public-health chair is yours after all. Each dot is a person; the lines are the contacts the infection travels; the bigger the dot, the more partners. Most of the dots are small — people with one or two partners in the window we are watching — while a handful swell toward the center with many. That lopsided spread is the whole point: a few hubs amid a crowd of ones and twos. Off in the dark beyond the frame, if you scroll to zoom out, you will find little closed clusters floating on their own — couples and small groups who are faithful to one another and never touch the wider network. They have almost nothing to do with how the disease spreads, but they are real people and they count in the totals, which is part of why the infected percentage runs lower than the crowded center alone would suggest. This is a small town’s worth of people, a few hundred; the individuals who turn up in the studies with partners in the hundreds are a lifetime tally across a whole city, not anything you would see in a snapshot this size, so read the biggest dots here as “the relatively most-connected,” not as literal centurions. The dots ringed in amber are the hard-to-reach reservoir — think of them as the unhoused, densely tied to one another. Scroll to zoom and drag to pan; the network spreads apart as you zoom in so you can trace who is tied to whom. Now your budget is measured in dollars, not doses. Treating a housed person is cheap and it sticks. Reaching a reservoir person costs several times as much, and only that expensive wraparound care actually suppresses them; treat them on the cheap and it barely registers. Pick a policy, set the budget, and watch the same outbreak on the same network respond. Hold the budget fixed and flip between policies — that is the only fair comparison, and the disease does not get a vote.

Things to try:

That experiment hides a second lesson underneath the first, and it is the one San Francisco is still living. The simple rule — treat the hubs — is right only while every dose costs the same. The moment treatment costs more for some people than others, the rule changes: you stop ranking by who is most connected and start ranking by how much suppression each dollar buys. A dollar of cheap pills on a housed hub buys near-total suppression. A dollar on the reservoir buys almost nothing, because the pills do not stick; it takes four dollars of wraparound care to do the job there. So the naive moves both fail in their own way. Spend only on the cheap hubs and you bend the wave hard, but the reservoir you skipped keeps a trickle of new infections going long after the main wave has passed, and the final toll creeps higher than it had to. Spend first on the expensive reservoir and you starve the cheap hubs and the wave rips through the main population. Only the policy that weighs leverage against cost — cheap hubs and wraparound for the reservoir, nothing wasted on the low-value middle — halts the spread soonest, leaving the fewest people infected in the end.

And this is why the last mile is the most expensive mile. Knocking the rate of new infections most of the way down is cheap and fast; extinguishing the last new infections costs more, per case prevented, than everything that came before, because those final cases keep flaring out of the reservoir, and the reservoir is precisely the part that money has the hardest time reaching. San Francisco drove new HIV infections down by more than seventy percent and then flattened out, chasing a zero it has not caught, and the people who account for the gap are disproportionately the unhoused — about a quarter of new diagnoses, suppressed at two-thirds the rate of everyone else. The uncomfortable structural truth, and it is structural before it is political, is that the cheapest road to zero runs straight through the most expensive-to-reach people. A society that will not pay to house and hold them does not save the money. It pays instead, indefinitely, in a disease that never quite goes away. The network keeps the reservoir warm no matter what anyone believes about whose fault it is.

There is one more door, and the network does not care which door you use. To the graph, a hub stops mattering the instant its edges go dark — and it makes no difference whether they were cut by a pill that renders the person non-infectious, by abstinence, or by a locked one. Quarantine is the third option, and on the bare mathematics it is the cleanest of the three: it needs no one’s cooperation, so it sails straight past the adherence problem that makes the reservoir so expensive, and it is cheap. Detain the hubs and the network shatters; the disease has nowhere left to travel. We already accept this exact logic in one corner of life: we lock up murderers and rapists against their will, not only to punish them but to protect the people they would otherwise harm, having decided the cost to their liberty is worth the safety it buys the rest of us. A disease super-spreader is, structurally, the same shape of threat — harm flowing through one person to others who never consented to the risk — and the structurally identical remedy is to take them out of circulation. Most free societies have looked hard at that argument and refused it anyway, reserving forced quarantine for a few extreme cases like untreatable tuberculosis, because they weigh the liberty of a person who has committed no crime differently from a convict’s. That refusal is a judgment about values, not about math. The network will tell you, to the decimal, how many infections a quarantine would prevent and how few people you would have to lock away to do it. It will not tell you whether you are allowed to. That line is ours to draw, and the arithmetic falls silent at exactly the point where the question starts to matter.

The reason the hubs win here is the reason Kevin Bacon can be connected to almost any working actor in three or four steps, the reason a handful of sites like Google and Wikipedia sit at the center of the entire web, the reason knocking out a few hub airports can ground a continent while closing a hundred small ones changes nothing anyone notices. Scale-free networks have a peculiar double nature. They shrug off random damage — delete a node at random and almost nothing happens, because almost every node is a leaf with a connection or two — but they are exquisitely vulnerable to a deliberate strike at the hubs. For the engineers who keep the internet running, that fragility is a security nightmare they spend careers guarding against. For an epidemic, the very same fragility is the cure. One structure, read once as a weakness and once as a lever, depending only on whether you are the attacker or the defender.

And the universe builds these networks over and over, anywhere that new things attach preferentially to whatever is already well attached — anywhere the rich get richer. New web pages link to sites that are already linked-to. New papers cite papers that are already cited. New actors want roles alongside actors who already work. New partners are drawn, disproportionately, to the people who already have many. Preferential attachment is one of the universe’s reliable recipes for structure, and every network it builds inherits the same soft spot. The disease found ours. The good news, the genuinely good news, is that we can use the same soft spot to find the disease.

The math does not care who deserves the medicine. It cannot be shamed and it cannot be flattered and it has never read a commandment. It knows only where the disease lives and which doses would break the network that carries it. If we can manage to want what works a little more than we want to be right, the tool is already in our hands. It has been there the whole time. We only had to be willing to point it at the people we had spent so long pointing fingers at instead.