May 25, 2021

Back in February 2020, a blogger named Jacob wrote a post on his blog, Put A Number On It, called “Seeing the Smoke”. The title comes from a famous 1968 psychology experiment in which groups of three students were put in a room to fill out a questionnaire. Then the experimenters started pumping smoke under the door and watched how the students reacted.

Two of the “students” in each group were paid actors, told to ignore the smoke. The remaining, real students, presumably not wanting to look stupid or panicky in front of the others, ignored the smoke in 90% of cases. If students were on their own, however, they went to investigate the smoke 75% of the time.1

The point is that we don’t just care about whether there’s a real chance of danger – even apparently real, apparently obvious signs of very imminent danger, like smoke under a door. We also care, very much, about looking silly.

Seeing the Smoke drew an analogy with the coronavirus, which was then still mainly in Wuhan. “Human intuition is … very good at one thing: not looking weird in front of your peers,” wrote Jacob. “It’s so good at this, in fact, that the desire to not look weird will override most incentives.” The point of his blog was to give permission to worry. If you were the sort of person who would sit in a smoke-filled room until someone else stood up, he said, “I’m here to be that someone for you.” The point was to give you social permission to start making preparations, whatever those preparations were.

I read Seeing the Smoke, and shared it on Twitter, in an attempt to spread the permission. Dominic Cummings, at the time still the chief advisor to Boris Johnson, did as well, and said so in March this year. It, along with a Slate Star Codex blog post a few days later, “helped me and some others in no10 realise we were going terribly wrong”, Cummings tweeted. Put A Number On It is a key part of the rationalist blogosphere, and Cummings, like me, is a fan.

Cummings is going in front of a Commons select committee on Wednesday, to talk about the decisions the Government made as the pandemic spread to the UK. There will probably be mud-slinging and a lot of claims and counter-claims about who said what and when.

Certainly, the arguments over whether or not the Government was “pursuing a herd immunity strategy” seem extremely vulnerable to definitional problems – does that mean they planned to let the virus rip unchecked, or to slow and control its spread? But it’s definitely the case that in early March, the UK was talking about slowing, rather than stopping, the virus: its pandemic flu plan said “It will not be possible to halt the spread of a new pandemic influenza virus, and it would be a waste of public health resources and capacity to attempt to do so.” Instead, they wanted to let the virus work its way through the population slowly, protecting the most vulnerable, while preventing it from ever overwhelming the NHS. You can call that a “herd immunity strategy” if you like, or not, but that was the goal.

And it does sound as though by 18 March, Cummings was pushing for faster lockdown. That was a few days after the publication of Seeing the Smoke, and around the same time as Tomas Pueyo published his influential post The Hammer and the Dance, calling for draconian, immediate lockdowns to crush the curve, which – if you acted early enough – could be reasonably brief, weeks rather than months.

Whatever one’s view of Cummings, an ability to see the smoke, despite the other people in the room apparently not seeing it, is an extremely useful skill. At the beginning of the year, the media didn’t want to look panicky, it didn’t want to look weird, so too many of us said things like “don’t worry about coronavirus, worry about the flu” or made fun of Silicon Valley types for refusing to shake hands.

There’s a phrase that is used a lot on social media, “read the room”. When people are angry about something, and someone else comes in and says “actually it’s a bit more complicated than that”, they are told to “read the room”. A classic example: when the political analyst David Shor tweeted, during last year’s Black Lives Matter protests, that violent race riots in 1968 reduced the Democratic vote enough to tip the election to Nixon, he was told to “read the room” and that he was “tone deaf”. A few days later he was fired from his job.

“Reading the room”, understanding what the social consensus around some issue is and not straying too far from that consensus, is an important thing. I’m not belittling it: we live in a social world, and we need to understand the mores and demands of that social world in order to prosper in it – and to make that social world bearable for everyone.

If we all ignored the strictures of society, it would be chaos. That includes rules of not saying things that are true but offensive: parents will know the exquisite agony of having their too-young-to-read-the-room child saying things like “Look daddy, that man’s really fat”. I’m not calling for a total free-for-all where you can say anything you like as long as it’s true. But reading the room isn’t always useful, and it won’t tell you if the room is filling up with smoke or not.

Back in March 2020, there was an elite consensus on certain things. Masks don’t work. We shouldn’t panic (37 days before lockdown, UK newspapers were mocking people who were scared of Covid and telling you you were more likely to be killed by a cow). Travel bans don’t work. (I got that one wrong.) If you locked down early, the population would get “behavioural fatigue” after a few weeks.

And if you had read the room, you would have agreed with all of it. I know I often did. That’s what the experts are telling us! It’s only Silicon Valley tech bros who want us to stop shaking hands, and Donald Trump who wants us to close our borders. The trouble is, the room was wrong, and reading it correctly didn’t help.

The rationalists, as Matt Yglesias wrote after the Slate Star Codex kerfuffle earlier this year, are really bad at reading the room, and much better at seeing the smoke. They talk about numbers and facts a lot, and come to conclusions like “actually, if you want to do good in the world, you should contribute to antimalarial charities rather than getting angry about the hot-button issues of the day”. That is the opposite of reading the room.

Cummings, for better or worse, is also terrible at reading the room. And he admires people who are similarly bad at it – the abovementioned David Shor, for instance. He promotes rationalist and rationalist-adjacent people like the podcaster Julia Galef and the superforecaster (and occasional UnHerd contributor) Michael Story. He wanted to hire “weirdos” as government advisers. (As an aside, I suspect that one’s ability to read the room is related to one’s tendency to “decouple”.)

These days, it’s wrong not to like masks, and we’re criticising the Government for not shutting borders quickly enough to counter the Indian variant. Behavioural fatigue turned out not to be a thing. I haven’t shaken hands in 14 months. But we’re still reading the room when we agree with those things. It’s just that the room’s opinion has changed.

An interesting game is trying to work out what the room’s opinions will shift on next. At the moment, it seems to be starting to change on the question: did the coronavirus escape from a lab?

It’s funny, because the hypothesis “Covid escaped from a Chinese lab” sounds like a wacky conspiracy theory, but all its constituent parts, “there are labs working on gain-of-function research in viruses, including in Wuhan”, “major disease outbreaks have been started by viruses leaking from high-security labs”, and “the Chinese aren’t always super open and transparent about things”, are pretty uncontroversial. Whether it turns out to be true or not, the theory has gone from crackpot speculation to plausible hypothesis, but very few of the facts have changed; only the opinion of the room.

And interestingly, Put A Number On It borrowed the “smoke” metaphor from an earlier blog post by a different rationalist, Eliezer Yudkowsky, who was talking about the dangers of human extinction caused by AI. Perhaps we ought to take that a bit more seriously too. (Although I should declare an interest, since my first book was on that topic.)

Seeing the smoke, not reading the room: these are difficult things to do. They involve putting yourself outside the consensus, outside the cosy circle where we all agree: they risk making you look ridiculous, or worse. In the Covid era, though, reading the room rather than seeing the smoke probably cost hundreds of thousands of lives. Dominic Cummings has many faults, but I suspect he is better at seeing smoke than he is at reading rooms.

FOOTNOTES
  1. A word of warning. This is a classic psychology experiment, and classic psychology experiments have had a bad time lately. A related concept, the “bystander effect”, is looking shaky. But I think this is one of the solid ones. Even if it isn’t, we’re only using it as an analogy.