Probability and Bayesian Inference
Why Your Brain Is a Bad Detective (And How to Fix It)
Every time you feel certain about something, there's a quiet mathematical error humming underneath that certainty.
The Idea
Most of us were taught to think about probability as a fixed fact about the world — a coin has a 50% chance of landing heads, full stop. But there's another way to think about probability, one that turns out to be far more useful and far more honest about what probability actually is: a measure of your current state of knowledge, not a property of the universe. This is the Bayesian view, named after the 18th-century English minister Thomas Bayes, and it does something radical. It says that when you get new evidence, you shouldn't just accept or reject a belief — you should update it proportionally. You start with a prior (your best guess before the evidence arrives), then revise it based on how well different hypotheses would have predicted what you just observed. The result is a belief that is never fully certain and never fully abandoned — it shifts, calibrated by evidence. This sounds obvious until you realise how rarely we actually do it. We tend to treat our first strong impression as a verdict, then recruit new information as a lawyer would — to support the case we've already decided. That's not reasoning. That's confirmation bias wearing the costume of reasoning. Bayesian thinking forces a different question: not 'Does this evidence prove my theory?' but 'How much more likely is my theory given this evidence, compared to rival theories?' That one shift — from binary to probabilistic, from verdict to calibration — changes almost everything about how you reason.
In the World
In 1999, a British woman named Sally Clark was convicted of murdering her two infant sons, both of whom had died of what might have been sudden infant death syndrome. A paediatrician testified that the chance of two SIDS deaths in the same family was roughly one in 73 million. The jury heard that number and convicted. The statistic was wrong in two compounding ways. First, it didn't account for the fact that SIDS risk runs in families — the deaths weren't independent events. But the deeper error was Bayesian. The prosecutor's logic implied: this outcome is very unlikely under innocence, therefore she's guilty. But that ignores the other side of the equation — how likely is it that a mother murders both children? That's also vanishingly rare. To reason correctly, you have to weigh both: how probable is this evidence if she's innocent, versus how probable if she's guilty? This fallacy — treating a low probability of the evidence under one hypothesis as proof of the alternative — is now called the Prosecutor's Fallacy, and it has distorted verdicts in courtrooms around the world. Sally Clark was eventually acquitted on appeal, but not before spending three years in prison. She never recovered from the ordeal and died in 2007. The tragedy isn't just personal. It's a demonstration of what happens when people in positions of authority reason about probability as if it were a binary on/off switch rather than a tool for comparing competing explanations.
Why It Matters
You don't need to be on a jury for this to be relevant. Every time you interpret a medical test result, assess a rumour about a friend, or decide whether a new piece of news confirms your political instincts, you are — whether you like it or not — doing probability. The question is whether you're doing it well. The Bayesian habit asks you to hold two things in tension: genuine openness to updating your beliefs, and genuine respect for what you already had good reason to think. Strong prior evidence doesn't disappear just because something surprising just happened. Equally, a surprising result shouldn't be explained away just because it disrupts a comfortable model. This isn't just intellectual hygiene. It's a disposition — one that makes you harder to manipulate, more honest with yourself about uncertainty, and better at changing your mind for the right reasons rather than the wrong ones. In a world engineered to produce outrage and conviction, calibrated uncertainty is quietly radical.
A Question to Ponder
What is one belief you hold with high confidence — and what would it actually take, in terms of evidence, to meaningfully shift it?
Get a new one of these every morning.
Start learning with Thinkable