Tom Chivers

Tom Chivers is a science writer. His first book, The AI Does Not Hate You is out now.


Dominic Cummings and I share some interests, to my slight chagrin. Both he and I are big fans of the nerdy, abstruse bit of the internet filled with people who are sometimes called “Rationalists”.

They’re a strange lot: smart and thoughtful, but strange. I wrote a book about them; they worry about AI destroying the world, but also about thinking accurately, about making good predictions, about doing the most good they can via charitable donations.

I know that Cummings is a fan, because he regularly quotes them. His blogroll includes links to LessWrong, Slate Star Codex and Eliezer Yudkowsky, three of the key parts of the Rationalsphere. His blog posts – like those of Scott Alexander, author of Slate Star Codex – are immensely long, although Alexander’s are clear and funny and designed to hold the reader’s hand through a complex argument, while Cummings’s tend to be a grab-bag of talking points thrown together with no discernible – to me, at least – structure. But a lot of the material, and the names mentioned, are similar.

Anyway. Yesterday, the Prime Minister, Boris Johnson, decided to suspend, or “prorogue”, parliament for five weeks around conference season. He and his outriders have argued that 1) it’s normal behaviour and that 2) it’s to push through his domestic agenda; but 1) it’s not (it’s the longest prorogation since 1945), and 2) it’s very obviously intended to reduce the amount of time available to backbenchers to legislate against no-deal or otherwise get in the Government’s way. BuzzFeed’s Alex Wickham reports that they have various other plans to further obstruct any parliamentary oversight.

It reminded me of something, and I was trying to work out what it was. And I realised: it was the classic application of game theory to the game of chicken. Imagine you’re playing chicken – that is, driving your car head-on towards another car. Whoever swerves first loses. What’s your best strategy?

One game-theoretic answer is: you ostentatiously unscrew your steering wheel and throw it out of the window.

On one level, it’s limiting your options, so it seems like a bad plan. But from your opponent’s point of view, it is a credible commitment that you are going to continue in a straight line. If the opponent wants to avoid disaster, they have to swerve, and lose the game. The example was first given by Bertrand Russell in his 1959 book Common Sense and Nuclear Warfare.

Russell thought it was a stupid idea. But the principle of credible commitment to a course of action formed a key part of the thinking of other game theorists, such as Herman Kahn, Thomas Schelling and John von Neumann, who were important in the development of the Mutually Assured Destruction doctrine – essentially, nuclear chicken. Ostentatiously commit to the idea that in the face of any aggression you will render much of the globe uninhabitable for millennia, and your opponent will not dare to be aggressive. Richard Nixon approvingly called it the “madman theory” and it is most associated with Schelling.

This is pure speculation, I should say – I have no special insight into Cummings’s mind – but I have a feeling this is what he’s doing here. I know he knows about this stuff: “Schelling points” and other game-theoretical terms are part of the jargon in the blogs we both read. Cummings has posts tagged “game theory” in his blog and discusses John von Neumann at length. If I’m right, Cummings has convinced Johnson to throw the Government’s steering wheel out of the window. By making it harder for Britain to change course from its headlong path towards no-deal, he thinks he can force the opponent to swerve.

I don’t want to buy into the widespread narrative that Cummings is some sort of all-seeing genius, though, if this is really what he’s doing. The reason the ostentatious removal of the steering wheel works in chicken is because you have one clearly defined opponent of a similar stature to you. If you’re both driving Vauxhall Zafiras, then the crash will be equally bad for both of you. The reason it worked, if indeed it did work, in the Cold War is because there were two obvious main actors, the US and the USSR, and nuclear annihilation would be comparably bad for both.

It’s not clear that the same is true in this case. For one thing, who’s the opponent? Is it the EU? Or is it Parliament? In Cummings’ and Johnson’s minds, it could be any one of them. Maybe they’re trying to force Parliament to vote through whatever deal he gives them, because the alternative will be a disastrous no deal and no time to fix it. Or maybe they’re trying to force the EU to make changes to the backstop because there’s no time for Parliament to block no deal. When you’ve got three cars in a game of chicken, it is much harder to model the outcomes.

More importantly: if it’s the EU, then it’s not two Vauxhall Zafiras. It’s more like a Fiat Cinquecento driving towards an oncoming train. A no-deal Brexit will hurt the EU, but much less than it would hurt Britain, because the EU is much bigger than Britain. And there are worse outcomes for the EU, such as other countries seeing that they blinked first and trying similar strategies to get favourable deals from them. It may be that no-deal is the least bad option available for them if Britain commits to it.

As I said: I’m speculating. But it fits my existing mental model of Dominic Cummings, which is that he reads these fascinating, interesting ideas from brilliant people, and takes completely the wrong message from them. For one thing, a key principle of the Rationalist blogosphere that we both admire is the “principle of charity”, the idea that “if you don’t understand how someone could possibly believe something as stupid as they do, that this is more likely a failure of understanding on your part than a failure of reason on theirs”, rather than assuming people are stupid or evil.

M’colleague Peter Franklin thinks that it’s unfair to accuse Cummings of having an “everyone-is-an-idiot-apart-from-me” attitude. That may be true, but he certainly doesn’t apply the principle of charity widely; he openly derides as a “courtier” the late former cabinet secretary Jeremy Heywood, for instance, who others have described as “the smartest guy in the room”. He may be right, but it seems sensible to at least consider the possibility that this much-admired figure had some understanding of how the machinery of government worked which Cummings was missing.

Another, arguably more relevant, principle is what is known as “Chesterton’s fence”. The conservative Christian GK Chesterton said that reformers often want to remove things because they don’t understand what they’re there for: he imagines one coming across a fence in a field, and pulling it down, because he can’t see the point in it. The reformer is then gored by a bull.

A wiser reformer, says Chesterton, would have said: “If you don’t see the use of it, I certainly won’t let you clear it away. Go away and think. Then, when you can come back and tell me that you do see the use of it, I may allow you to destroy it.” For Chesterton, the strange and sometimes incomprehensible workings of tradition are probably there for a reason, and if you start pulling on threads because they don’t look important, the whole thing might come apart.

Cummings may work for the Conservative Party, but he is not a conservative, in the Chestertonian sense. He’s willing to erode the conventions of the British constitution, by suspending parliament for political reasons and trying to keep them from voting on a huge and important issue.

He’s got form. In 2010 he and Michael Gove forced through legislation on academy schools by fast-tracking it in a way usually reserved for emergency laws, things like counter-terrorism. Chesterton, and the “Rationalists” Cummings admires, would be deeply wary of eating away at the various little conventions and traditions which underpin British public life in the pursuit of a short-term goal.

My concern – one of my concerns – is that Cummings will come to be seen as representative of the Rationalists, because he quotes them and shares some of their concerns and interests. But at the risk of being uncharitable myself, I don’t think he really understands them. He thinks it’s all about seeing how everyone else is wrong, and applying game theory to politics, and all that. And those things are important.

But it’s also about looking at your own thinking, and seeing where you are going wrong, and taking steps to make sure that you’re not going to accidentally blow everything up with your brilliant ideas. I worry that Cummings isn’t playing some clever John von Neumannesque game, but driving his Fiat down the rails and expecting a train to swerve.

Tom Chivers’s book, The AI does not hate you, is out now