Be Careful
What Margin Call Got Right, What I Lived, and Why Bias Is Now a Public Health Crisis
There is a scene in Margin Call that most people remember for the wrong reason.
Jeremy Irons, playing the chief executive of a fictional investment bank on the night its risk model breaks, sits at the head of a boardroom table and asks Zachary Quinto’s junior analyst to explain what he has found. “Speak,” he says, “as you would to a young child. Or a golden retriever.”
The standard reading of that line is condescension; the powerful man performing ignorance to humble the technician. That reading is wrong.
Irons is not performing ignorance. He is performing something far more disciplined and far more rare. He is demanding that the model be explained without the protective coating of jargon, without the institutional scaffolding that allows complex things to sound reassuring, without the vocabulary that intelligent people use to avoid saying what they actually mean.
He wants to know if the person in front of him actually understands what they have built.
That is not the behaviour of a man who cannot follow the analysis. It is the behaviour of a man who has been in enough rooms, over enough years, to know that the most dangerous thing in any institution is not the bad model. It is the model that sounds right. The one that nobody challenges because challenging it would require someone to say out loud what it actually shows.
I know that room. I have sat in it. Not in a film. In a building in North London and then Bingley, in 2005, explaining to executives what our models were showing about the mortgage book we had built.
This article is about what happened in that room. What always happens in that room. And why, in 2025, with every tool, every dataset, every LLM and every published methodology freely available, the people who should know better are still choosing not to model. Not because they can’t. Because the model would threaten something they are attached to.
