A "bias," in the everyday sense, sounds like something you'd want to fix. A cognitive bias is closer to the opposite — it's what your judgment does when it's working exactly as designed, trading a bit of accuracy for a lot of speed. That trade made sense for most of human history, when a fast wrong guess about a rustling bush beat a slow correct one. It makes a lot less sense when the "bush" is a headline, a price tag, or a stranger's argument.
A heuristic is a shortcut, not a mistake
Amos Tversky and Daniel Kahneman spent decades cataloguing these shortcuts, and the useful thing about their work is that it separates the shortcut from the error. The availability heuristic judges how common or likely something is by how easily examples come to mind — which works fine most of the time, and fails specifically when the vivid case and the common case come apart. Plane crashes make the news; fatal car accidents rarely do. The heuristic hasn't broken. It's just optimising for "what's memorable," and memorable was never the same thing as "frequent."
The anchoring effect is the shortcut of leaning on the first number you saw, even when you know it's arbitrary. Tversky and Kahneman spun a wheel rigged to land on either 10 or 65, asked people to guess what percentage of UN countries were African, and the group that saw 65 guessed higher than the group that saw 10 — despite the wheel obviously having nothing to do with Africa. Nobody in that room was being careless. The number was just already in their head, and a mind under time pressure builds its estimate near whatever's already there.
Confirmation bias is the one that compounds
Most heuristics misfire in isolated moments. Confirmation bias misfires continuously, because it shapes what evidence you go looking for in the first place — a mind that's already leaning toward a conclusion notices the facts that support it and quietly stops looking once it's found enough of them. It's not that contrary evidence gets rejected; often it never gets searched for at all. That's what makes it the hardest of the three to catch in yourself: the distortion happens before the disagreement does.
What we're still unsure about
Naming a bias is not the same as neutralising it. There's a well-documented gap between "I can define anchoring" and "I don't anchor" — knowing the mechanism intellectually and having it stop influencing your actual guesses are two different skills, and the second one is much harder to train than the first. Debiasing training has a mixed track record for exactly this reason. We built the curriculum's Cognitive Psychology domain because recognising the pattern by name is still worth something — just not everything.