How Do We Know That We Know?

There’s a saying that gets passed down to medical students like a family heirloom. Some dean stands up during orientation and tells the new class: half of what we teach you here will turn out to be wrong. The problem is that nobody knows which half.

I don’t remember anymore whether anyone said it at mine. It doesn’t matter, because medicine spends the rest of your training proving it right. Treatments I memorized for exams have since been demoted. Thresholds moved. Diseases got reclassified. The textbooks I studied from have editions the way phones have models, and each edition quietly deletes things the previous one stated as fact.

The saying is usually delivered as a joke about humility. Taken seriously, it’s a bomb. If we can’t tell the right half from the wrong half, what exactly do we mean when we say we know something?

On Monday I argued that your senses never show you reality, only a translation of it. A reader named Randy ( I love all the comments) put it better than I did in the comments: the translation is “optimized for action, not truth.” Which raises the obvious next question. Fine, perception is a translation. But surely the things built above perception, the theories and proofs and diagnoses and laws of physics, can be trusted in a way the raw senses can’t. Reason should succeed where the eyes fail.

Today I want to show you how deep this well actually goes. The branch of philosophy living at the bottom of it is called epistemology, and it circles one question I’ve been chewing on since long before this newsletter existed: how do we know that we know?

Start with the most basic thing you could possibly ask. What is knowledge?

For roughly two thousand four hundred years, philosophy had an answer that looked airtight, and it traces back to Plato. Knowledge is justified true belief. Three conditions. You believe it. It’s actually true. And you have good reasons for believing it, so you’re not just guessing. If it rained last night and you believe it rained because you watched it through your window, congratulations, you know it rained. Belief, truth, justification. All three boxes ticked.

Plato himself, in the dialogue where the idea appears, wasn’t fully convinced by it. The formula outlived his doubts anyway, and for most of Western history that was simply what knowledge meant.

Then in 1963, a philosopher almost nobody had heard of, Edmund Gettier, published a paper three pages long and broke it.

The cleanest version of the problem was actually described by Bertrand Russell years before Gettier formalized it. A man glances up at a clock. It says two o’clock. It really is two o’clock. Clocks are reliable, so his belief is justified. Justified, true, believed. By the classical definition, he knows the time.

Except the clock is dead. It stopped exactly twelve hours earlier, and he happened to glance up during one of the two minutes a day when a dead clock tells the truth.

Something in you refuses to call that knowledge. He was right by pure luck, and luck is precisely what the justification condition was supposed to eliminate. So there is a fourth ingredient, something beyond belief, truth, and reasons, and the genuinely uncomfortable part is this: sixty years later, philosophers still haven’t agreed on what it is. Every proposed patch springs a leak somewhere else. Which leaves us in a strange position. The word every courtroom, exam, and lab report depends on has no accepted definition. We use it fluently. We cannot say what it is.

That’s the small problem.

The big one is older, and it aims at science directly. The engine of all science is a move called induction. The sun has risen every morning in recorded history, therefore it will rise tomorrow. This drug lowered mortality in ten thousand patients, therefore it will lower yours. Every measured constant, every dose curve, every law of physics is a record of the past, extended into the future.

David Hume, writing in the 1740s, asked the question that has never been answered: what justifies that extension?

Try it yourself, and notice that every answer cheats. “Nature is uniform.” How do you know? “Because it always has been.” That is induction testifying on behalf of induction, a witness vouching for himself. There is no independent ground. The one move science cannot function without has no non-circular justification, and Hume understood exactly how radical that was. By his own account he coped with it by having dinner and playing backgammon with his friends, then returning to work the next morning as if the abyss weren’t there.

To be clear, none of this is an argument to stop trusting the sun, or your medications. Planes fly. Antibiotics work. The point is to notice what kind of trust this is, because it is not the kind you thought you had.

And underneath both of these problems sits the oldest one of all. An argument attributed to a Greek skeptic named Agrippa, some two thousand years ago. It is so simple that every child performs it naturally, and so brutal that in two millennia nobody has escaped it. Take anything you believe. Ask why. Then ask why again, and keep going, the way a four year old does, and every chain of justification a human being can construct will end in one of exactly three places.

All three should bother you.

Behind the paywall: the three dead ends, the single tile of certain floor Descartes found at the bottom of them, the move Kant made that connects straight back to Monday’s post, and then the part I care about most, which is how medicine taught me to act on knowledge nobody can prove, and why I’ve made a strange kind of peace with that. If you’ve been thinking about upgrading, it’s free to try.