The Balloon Debate

There used to be a room built specifically to make this happen to you.

You sit across from someone whose only job, for the next hour, is to find the hole in what you've just said. Not the facts — you were expected to have those before you walked in. The argument. The thing you built out of the facts. You defend it, out loud, with no one else in the room to hand the question to, and if it doesn't hold, everyone present watches it not hold, starting with you.

That was more or less the tutorial system British universities built themselves around for a very long time. Oxford and Cambridge get the reputation, because one tutor and one terrified undergraduate in a room photographs well, but the same exposure turns up across the older universities generally: the viva you stood up for at finals, the supervision where someone paid to disbelieve you took the essay apart line by line, the seminar where there was nowhere at the back to hide. Ask around and most people who went through any version of it, at any university, will tell you the same thing without needing the reference explained — they hated it at the time, and they can still describe the room forty years later. It's also, not coincidentally, why the wit of someone like Oscar Wilde reads as something closer to a discipline than a talent — timing, bias, social pressure, the exact temperature of a room, judged and answered for in real time, in front of the people it was aimed at, with nowhere to take it back. E.M. Forster's characters keep facing the same demand in miniature: not what do you know, but what will you decide, and will you stand there once you have. Wellington didn't get to convene a working group at Waterloo. Nelson didn't loop in stakeholders before he raised the telescope to his blind eye at Copenhagen. Napoleon's genius and Napoleon's ruin were produced by the same faculty, running without a safety net, and both were entirely his. None of it was reviewable. The judgement and the consequence lived in the same body, at the same moment, and everyone could see whose body it was.

I don't think we built a worse version of that room on purpose. I think we built an enormous, well-meaning machine for making sure it almost never has to happen to anyone again.

It wasn't my call. Four words, and they might be the most useful sentence in modern working life — useful precisely because they're almost always at least partly true. Somebody else set the policy. Somebody else approved the process. Somebody else, three approvals up the chain, will make the decision that actually counts. Multiply that sentence by every layer of an organisation and you get the working life most of us actually live in: the email you send because your manager asked you to send it, the report you don't fully trust, referred back to IT rather than corrected by you. Nobody in that chain is being dishonest. Every single handoff is the locally reasonable choice. Add enough of them together, though, and you get a system almost perfectly engineered so that when something goes wrong, no one is quite standing where Wellington stood, whether he liked it or not.

I know what standing there actually feels like, because I did it once, and it took more than a moment in a room to prove. I was headhunted by Goldman Sachs earlier in my career and offered the job in the first interview — no HR round, nothing. It didn't feel right, and not in any way I could have written down as a reason. The offer was full of planned holiday perks and business obligations that would sit inside every part of my life, not just the working part of it. It read less like a salary and more like a lifestyle they intended to run. I said no on a Thursday.

Saying it once turned out to be the easy part. I left for a week in Cornwall the next day. They rang the office the Friday before I went, and then they contacted me every single day I was away. I came home and told them again: no. A few days later their external headhunter asked to meet. We had a glass of wine in a bar. She'd worked at Goldman herself, and she told me, kindly, that people sometimes sat seventeen interviews just to move internally, that she'd never once heard of anyone offered a role without even meeting HR, that I should be sure I understood exactly what I was turning down.

None of it changed anything. By every external measure in the room, the offer was extraordinary, and the person telling me so had spent a career learning to read precisely how extraordinary. My answer wasn't running on that measure. It was running on something the interview process had no column for — not what the job paid or how rare it was, but what would actually happen to my life, quietly, one convenient perk at a time, if I let someone else plan my holidays for me. Nobody handed me that answer. Nobody could have. It got tested daily for a week, and it held.

We didn't remove the mistakes. We removed the exposure. And it turns out the exposure was most of what was training anybody.

This is about to matter more, not less.

Whatever else the AI era turns out to mean, it means the part school has always been best at — holding facts, retrieving them on demand, certifying that you have them — is the part a machine will shortly do better than any human being alive, for free, in seconds. That was never really in question. What's left, once the facts are free, is the part underneath them: what you decide when the facts run out, or contradict each other, or arrive too late to help. School has an entire architecture for the first thing. For the second, it has almost nothing at all — not because anyone decided reasoning didn't matter, but because it's expensive to teach, hard to grade, and, until recently, not the part of the job that seemed urgent.

I don't think the fix is a new subject on the timetable. I don't think you redesign a curriculum built for scale and certification into something that trains exposure instead — that's asking the tunnel to become the mountain. I think you build the room back in on the side, cheaply, the way it used to exist almost by accident: the balloon debate.

Put a child in front of the class and give them a character to defend — a historical figure, a position, an idea that may have nothing to do with what they actually believe — and tell them the balloon is going down and only one passenger can stay in it. Make them argue it, out loud, on their feet, against people actively trying to throw their character overboard. Nobody researched this for homework. Nobody gets to bounce it to an expert. You turn over an argument that isn't yours, in public, and find out in real time what in it actually holds.

That's not a curriculum. It's barely an afternoon. But it's the closest thing most schools already have, almost by accident, to the room Wellington and Wilde and every Oxbridge undergraduate never got to leave — and it's sitting there half-forgotten as a novelty exercise, right when we're about to need it more than we have in a hundred years.