Watch a three-year-old for an afternoon and you are watching a human being learn faster than they ever will again. They point at the dog, you say "dog," and that is its name now, settled, no questions asked. They point at the moon, you tell them it is very far away and that people once walked on it, and they accept both the distance and the walking, though they have measured neither. They ask why the sky is blue, you give some half-answer you are not sure of yourself, and they believe it. They learn their own name because you told it to them. They learn their birthday — the one fact that sits at the very centre of who they are — only because someone said a date aloud and they took it.
A child believes everything. We tend to describe this as a charming weakness, something they will grow out of, an early and unfinished version of a mind that has not yet learned to check. But pause on it, because there is something odd in it. A child who believes everything, who takes every word as true, is not failing to know. That child is doing the one thing that has ever worked to bring a human being from nothing to a mind at all. Every word we have, every fact we hold, everything we know about the world, began in a small person who simply believed what they were told — because the alternative was to demand evidence, in a language they did not yet have, for facts they had no way to check, and so to learn nothing and stay empty.
The question for tonight is this. Suppose that is not the early, unfinished version of knowing. Suppose the child who believes everything is closer to the truth about how any of us knows anything than the careful adult we admired last time — the adult who weighs each claim against the evidence before accepting it. Suppose that believing first, and checking only sometimes, is not something we grow out of but something we never actually leave.
This is Philosophy for Us — philosophy for everyone, no degree required. This is the third of eight episodes on how you know anything you did not see for yourself: on deepfakes, on trust, and on what is left of knowing when the recordings themselves can lie.
Last time I set out a case as strongly as I could and gave it no reply. It was David Hume's. A wise person, Hume held, proportions belief to the evidence; another person's word is one more piece of evidence, worth no more than its record will bear; the burden is on the claim to earn its place, and until the evidence arrives the right response is "not yet." I said it might well be right. And I planted a single question in it and then left the question alone — not whether the rule is true, but whether it can be lived. Whether a human being could pass a single day believing nothing they had not checked for themselves.
Tonight I keep the promise I made then. I said I would bring a philosopher who takes the other side of that exact question — who holds that Hume did not merely get a detail wrong but mistook what a human knower is, and that the rule of verify-first describes no creature who has ever lived. He argues that the child is nearer the truth than the auditor. And, like Hume, he is no lightweight.
His name is Thomas Reid. He was Scottish, like Hume, and his contemporary — the two lived at the same time in the same small country, arguing the same questions, and for much of his life Reid was the most serious answer Hume ever received. He set out his reply in An Inquiry into the Human Mind on the Principles of Common Sense, published in 1764. Where Hume is cool and sceptical, and takes pleasure in showing how little our confidence is really worth, Reid turns and says: wait. If a philosophy ends by telling me I cannot know the most ordinary things — that I should not trust my own eyes, or my own memory, or the people around me, without a proof I can never actually supply — then perhaps what has gone wrong is the philosophy, not the knowing. Perhaps the argument is elegant and proves something that everyone, the philosopher included, knows to be false the moment they leave the study.
His reply to Hume on testimony — on the whole question of taking another person's word — begins from one observation, simple almost to the point of embarrassment, from which the rest of the view follows. It is this: look at how human beings actually behave with each other's words. Not how a philosopher says they should behave. How they do.
What we find, Reid says, is two things. He calls them principles, and he means the word strongly — not habits we picked up or rules we worked out, but part of the basic equipment, built in, as the eye is built to see. The first he calls the principle of veracity. We are built to tell the truth. Not perfectly and not always — people lie, plainly — but lying is the effort, the exception, the thing one has to work up to and then cover. The resting state, the thing that costs nothing, is to say what is so. It is hard to teach a small child to lie convincingly; the truth simply falls out of them. Speaking the truth, Reid says, is the default condition of the human animal, and lying is the departure from it. There is a reason we ask a liar "what came over you" — we treat the lie as the aberration.
The second principle matters most for tonight, and it carries an unfortunate name. He calls it credulity. To our ear the word is already an insult: a credulous person is a dupe, a mark, someone who will believe anything. Set that reaction aside for a moment, because Reid is using the word for something he takes to be among the most important features a knower can have. By credulity he means this: just as we are built to speak the truth, we are built to believe what we are told. The default, the resting state, the thing that costs no effort, is to take the word as true. Disbelief is the effort. Doubt is what has to be worked up to. A child does not decide to believe you; belief is simply what happens when you speak and they hear. They have to learn, slowly and late and never completely, to hold a word back and weigh it. Believing comes first. Checking is added afterwards — partial, expensive, and rare.
Put the two principles together and consider what Reid has built. We are disposed to speak the truth, and disposed to believe what is spoken. The two fit together: one part of humanity producing true reports by default, the other taking them up by default. That fit, Reid says — that matched pair of dispositions — is the actual source of everything the species has ever known. It is how the child acquires language. It is how language carries knowledge. It is how anything any human being has learned was passed to the next person who needed it. Remove the credulity — require every person to demand independent grounds before believing any word — and the process never starts, because the child has nothing to reason from, and so learns nothing, and grows up with nothing to check anything against.
So this is Reid's case against the careful auditor of last time, and it is a large one. Hume said belief must be earned, word by word, by evidence; the default is doubt, and trust is granted only once the grounds are in. Reid replies that this is exactly backwards — that it is not how any human being who has ever lived actually came to know anything. The default is not doubt. The default is belief. We arrive already believing, and we trim that believing back, here and there, where we have a specific reason to. The auditor weighing each claim and granting belief only where the evidence licenses it is not, Reid says, the foundation of knowing but a refinement built on top of it. The checking is real and worth doing; it simply comes second, and late, and only for some things. The believing came first, and most of it is never revisited at all. Hume, on Reid's account, described the rare exception and mistook it for the rule.
Let me make this precise, because the whole collision with last time rests on one exact difference.
Last time, with Hume, I described belief as a quantity that can be set at any level — not simply on or off, but adjustable, a little or a lot, and adjustable only by evidence. And where does it stand before any evidence arrives? At zero. That is the spirit of the view. You begin at not yet, at no, with the burden resting on the claim, and every degree of belief has to be earned with grounds. A stranger's word, arriving with no record behind it, earns almost nothing, because it has not yet earned the right to. That is Hume's account, rigorous and honest, and last time I said it might be right.
Reid offers a different account, and it is the one I want held clearly tonight. Belief does not begin at zero and get purchased upward by evidence. It begins as the default, and is given up only for cause.
When someone tells you something, on Reid's account, you do not start from no and let their evidence raise you towards yes. You start from yes. Belief is the default condition, the resting state, what happens unless something stops it. The word comes in and you believe it — not because you have weighed it and found the grounds sufficient, but because believing the spoken word is simply what a human knower does, as seeing the colour is simply what an open eye does. You do not earn your way into belief. You are already in it, and the only question is whether something will come along to put you out.
And something can. This is what keeps Reid from foolishness, and it is what the word "credulity" hides. The default does not mean you believe forever, helplessly, whatever happens. It means belief holds until a specific reason to doubt appears — an actual, particular reason, about this speaker or this claim, to withdraw it. The man has lied to you before. The account contradicts something you saw with your own eyes. He is slurring and swaying and trying to sell you something. He has every motive to deceive you and no fear of being caught. Any of these, and belief is rightly withdrawn. Reid is not telling you to believe the confidence trickster. He is telling you where the burden lies. For Hume, the burden is on the word to earn belief, and until it does, the answer is no. For Reid, the burden is on the reason to doubt to earn disbelief, and until one appears, the answer is yes. Belief is granted from the start, and you need a reason to withdraw it — not a reason to grant it, because it was granted all along.
Everything turns on that one reversal; it is the whole dispute. The same people, the same words, the same world — and the two philosophers place the burden on opposite sides. Hume: prove it before I believe it. Reid: I believe it unless you give me cause not to. And this is why Reid is the rival that the careful, do-your-own-research listener resists most, the one that unsettles them, because he tells that listener the thing they least want to hear. The stance they are proudest of — taking no one's word without grounds, making every claim earn its place — describes exactly one case: the single claim they are actively holding under scrutiny, while the thousands of other things they believe, and never think about, were all taken on trust without a second thought.
That is Reid's deepest claim, and it is a genuine argument, not merely a warmer mood than Hume's. He is not saying that default trust would be pleasant, or kind, or good for society. He is saying that default trust is what you are already doing — everywhere, all the time, in almost everything you hold — and that a picture of knowing which treats each belief as earned one item at a time has simply lost sight of how much of your believing was never earned and never could be, for want of the hours.
Now, and this matters, because I am not in the business of building Hume up one week and knocking him down the next: Hume has a reply, and a good one. He can grant every word Reid has said about the child and concede nothing. Yes, he says, the child must begin by believing; you cannot check before you have anything to check with. But that is a point about how knowing gets started, not about when a belief is earned. The child's early credulity is a loan, not a proof — an advance extended against the day when reason has grown enough to go back and examine the accounts. That we cannot, in one short life, examine every entry does not make checking a luxury; it makes checking the standard we always fall short of — and falling short of a standard is not the same as the standard being wrong. Reid has described how we come to believe. Hume was describing when we have a right to believe. Two different questions — and, Hume would say, Reid has answered the easy one and presented it as the hard one.
So you now hold two complete accounts, and they cannot both be the more basic one. Either belief begins closed and is earned open — Hume — or it begins open and is closed for cause — Reid. Either the burden is on the word or the burden is on the reason to doubt. Either the careful checker is the model of a knower and the trusting child is what we grow out of, or the trusting child is the source of all knowing and the checker is a late addition that rests upon it. These do not blend. You cannot keep both. One of them is the truth about what a knower is, and the other is a story we tell about ourselves that does not survive close inspection. And I am not, tonight, going to say which.
I want to slow down on one word, because a piece of history is hidden inside it that changes how the whole dispute looks, and most people never notice it. The word is trust. And I want to claim something that may sound like a stretch and is not: part of the reason the deepfake era feels like the end of trust is that we are using a narrowed version of this word, and we do not know it has narrowed.
Ask what trust means to you now. For most of us, raised when we were raised, trusting someone has come to mean roughly this: estimating, from the evidence, how likely they are to come through. I trust my mechanic — meaning his record is good, he has not cheated me yet, the odds that he is straight with me are high. Trust as a kind of forecast; a bet already priced. We speak of earning trust, building trust, losing trust, much as we would speak of a credit score — a figure that rises with good conduct and falls with bad, a running estimate of reliability accumulated over time. You can hear the whole apparatus in the phrases: a trust score, a trusted source, reputation treated as data. To trust, in this modern sense, is just to have assessed someone and arrived at a high number.
Go back a few hundred years and the word does not mean that. It means something larger and stranger, and more like an act than a calculation. To trust someone was to give them credit — and that word, credit, comes directly from the Latin credere, to believe, the same root that gives us creed, the thing believed, and credo, I believe. To give someone credit was to commit something of your own into their keeping, to depend on them before you had any proof they would hold. It was a relationship, and an act of the will, not a sum you had finished working out. You did not trust someone because you had calculated that they were safe. You trusted them, and in trusting placed yourself within their reach, and so made yourself liable to be let down. The wrong that answered that older trust was not miscalculation, not "I estimated wrongly." It was betrayal.
And notice how Reid's credulity is the same thing, seen from the side of knowledge. The default disposition to believe, to give the spoken word credit, to take what a speaker says into your keeping before any check — that is credere, that is giving credit; and giving credit is just what trusting someone amounted to, in practice, before the word narrowed. Reid is not inventing some optional, old-fashioned kindness. He is describing what trust used to mean — the older, larger sense of the word, before it contracted.
And contract it did, and you can almost watch it happen. Under the very pressure Hume applied — proportion belief to the evidence, treat every word as a data point, admit nothing unearned — trust was redescribed, without anyone deciding to. The relationship fell away, and what remained was the forecast. The one who trusts stopped being someone who commits himself and became someone who assesses. The one trusted stopped being someone depended upon and became a mechanism with a success rate. And it is the narrowed version we have inherited. We say trust and we mean the score.
Keep that in view, and look again at why the fakes feel catastrophic. If trust simply is verified reliability — if to trust is nothing more than to hold a good, evidence-based estimate of how likely a source is to be straight — then a technology that makes the evidence unreliable really does abolish trust. Corrupt the record, produce convincing fakes in overwhelming numbers, and the estimate becomes impossible, the score cannot be computed, and trust, on this meaning of the word, ends. That is the dread from the first night, and now its root is visible. It is not only that the cameras have begun to lie. It is that we had already, long before the first deepfake, exchanged a word that named a relationship for a word that names a calculation — so that when the calculation became impossible, it felt as though the relationship were dying, when the relationship may be another thing entirely, one the fakes do not touch. May be. I am not going to tell you tonight that it is, or where; that is a later night's work.
What I will say is that the older meaning is not gone. We still live by it, in the places that matter most, where no one computes anything. You did not trust the person who raised you because their reliability score passed a threshold. You were handed to them, helpless, and you depended on them entirely before you could so much as name them — and that was the most complete trust you will ever give anyone, and it held not one ounce of calculation. On the score-meaning of trust, that was irrational: belief wildly out of proportion to the evidence. Reid would call it the truest thing you ever did, and the model for the rest. Which sharpens the collision with Hume further, because now it is not only about belief. It is about whether the most basic thing we do with one another is a calculation we perform or a reliance we place.
Let me gather what you are holding, and then leave you with the part I am not going to settle.
You came in last time with a reflex — the pull that wants the proof before it will believe — and you left with that reflex made into a philosophy, Hume's, given whole, with no reply. Tonight I brought the reply, and not from me: from Thomas Reid, every bit Hume's equal, who holds that the careful checker has it backwards. On Reid's account belief does not begin at zero and rise as it is earned; it begins as the default and is withdrawn only for cause. We are built to speak the truth and built to believe what is spoken, and that matched pair, he argues, is not the soft thing we grow out of but the very thing that carried each of us from an empty-headed child to a mind that could so much as think of checking. And Hume, as you saw, has his reply ready: this is an account of where knowing begins, not of when a belief is earned. The dispute is live, and I am leaving it live. But notice what merely raising it has done to one quiet word. What Reid calls trust is older and larger than the score we have narrowed it to — it is credit, a commitment of oneself, the thing you gave whole, before you could count, to whoever first kept you alive.
So this is where you now stand. You have two accounts. Both were built by serious people who prized rigour. Both are coherent. And they point in opposite directions. Hume: prove it before you believe it; the trusting child is what we outgrow. Reid: believe it unless you are given cause not to; the trusting child is what we are. And you cannot escape the choice by being reasonable and splitting the difference, because they disagree about which of the two is more basic — about which one is the exception that rests on the other. There is no middle position between "trust is a refinement that depends on checking" and "checking is a refinement that depends on trust." One is basic, and the other rests upon it.
Now, you might think the way forward is obvious: find a third philosopher, cleverer than both, who settles it. And I could do that — there are people who have spent careers trying to reconcile these two positions, and we will meet some of them before we are done. But I am not going to reach for them next, and here is why. We have just watched two of the finest minds the subject has produced look at the same human being and describe two incompatible creatures. Adding a third description to the pile does not break the tie. It only gives you three theories to weigh instead of two, with the same judgement you do not yet know whether to trust. We have surveyed enough. The theories have done what theories can do, which is to lay out the live possibilities without dilution. They cannot tell you which one is you.
So next time we stop surveying, and do the thing I have been circling since the first night without yet letting us do it. We stop asking which philosopher is right about the human knower in general, and turn the question on the only knower you can examine from the inside — yourself. Not which theory is the better theory, but which of these two creatures you have actually been. And we settle it the one way it can honestly be settled: not by taking Reid's word, or Hume's, or mine, but by looking.
I am not going to tell you tonight what we will find; that would spoil the only part of this that does any good, which is your doing it and seeing it for yourself. Hume may be right. You may turn and look and find your beliefs checked, grounded, earned, as the careful picture says. Or you may find something else. I am genuinely handing you the question, not the answer: which of the two are you?
So carry this out with you. The next time you catch yourself proud of being a checker, a verifier, someone who takes no one's word without grounds, do not argue with the feeling. Put one question to it: how many of the things I know did I actually check? Not how many I could have checked, in principle, given the years — how many I did. Do not answer it yet. Hold it, as you held the last one, as an open question. Next time, instead of asking it in the abstract, we turn and look — together.
Thanks for listening. I'll see you next time.