A few years before any of this, I sat in a doctor's office and could not get my account taken seriously.
I had had a pain in my side for weeks. Nothing dramatic — just there, every day, in the same place. I had put off going, as one does, until I could not put it off any longer. So I went, I sat down, and I described it as plainly as I could: where it was, how long it had lasted, what made it worse.
And I could see the doctor reach a conclusion about me before I had finished the second sentence. I cannot prove that; one never quite can. But I had come in tired and anxious — I had been carrying this for a month and had talked myself into three different alarming explanations on the drive over — and I think she took all of that into account before she took the pain into account. Stressed. Run-down. The sort who reads too much and worries too much. Let us see how you are in a few weeks. Try to get some rest.
Here is the part I keep returning to. She may have been playing the odds correctly. Most people who arrive as I arrived — tired, worried, with nothing obvious to find — are in fact fine; a good doctor knows that, and rest really is the right advice most of the time. So I cannot honestly say she did something wrong. What I can report is how it felt from where I sat. It felt as though my account of my own body had been marked down before the evidence was in — not because of what I had said, but because of who I seemed to be when I said it.
It turned out to be real, as it happens. It took another month and a second doctor, but there was something there, it needed treating, and it would have been easier to treat a month earlier.
This is Philosophy for Us — philosophy for everyone, no degree required.
This is the third of nine episodes. Last week I did what I had promised: I took the answer most of us begin with — that our sense of who is credible is simply a fair record of who tends to be right — and I built it up to full strength. On that view, credibility answers to one thing only — accuracy; our judgments of whom to believe track how reliable people have proved to be; and even the view's sharpest critic grants that not every low judgment is a wrong. I built it to be convincing, and I left it standing.
Tonight I test it for the first time. Not by refuting it — it will still be standing at the end of the episode, exactly as it is now — but by setting one thing beside it that it cannot settle on its own. There is a kind of low judgment we hand out that is not tracking how reliable a person is at all. It is tracking who they are. And from the inside, that kind of low judgment and an honest one are indistinguishable. The question tonight is whether our fair, well-calibrated judgment can tell the two apart. Not whether it should care. Whether it can.
There is a philosopher who gave that feeling a name, and once one has the name the thing becomes visible everywhere. She is Miranda Fricker, and we met her briefly last week: hers was the book that made the concession the calibration view leaned on, the point that not every low judgment is an injustice. Tonight we meet the rest of her argument. The book is Epistemic Injustice, published in 2007, and the wrong it describes is the one I met in that office.
Here is the wrong, as plainly as I can state it. Sometimes a person tells us something, and we give their word less weight than it deserves — not because of anything to do with how reliable they are, but because of who they are. Their accent. Their age. Their job. Their colour. The fact that they are a woman in a room of men, or the child in a room of adults, or the patient rather than the doctor. Something about their identity reaches us before their words do, and lowers the weight we give them before they have had the chance to be right or wrong. Fricker calls this testimonial injustice. And its centre is this: the person is wronged specifically as a knower — as a source of knowledge. We have not merely been unkind to them. We have discounted their standing to tell us something true.
Notice how specific that is, because it is easy to blur it into ordinary unfairness and lose what is sharp about it. If I am rude to you, that is a wrong, but it is not this one. If I push past you in a queue, that is a wrong, but it is not this one. Testimonial injustice is a wrong done to a person in their capacity to know things and say them. It treats them as a less credible giver of truth than they are. And for a creature like us, which lives almost entirely on being able to tell others what it has seen and have that count, this is a deep harm. To not be believed about your own pain. To name a danger and see it dismissed by the people who could act on it. To be the one person in the room who actually saw, and the one whose word counts least.
And the source of it, Fricker says, is prejudice — though I want to be careful with that word, because what she means is broader and quieter than the version one first pictures. One pictures someone who means it: the bigot who looks at the speaker and decides, with feeling, to dismiss them. That happens, but it is not the main case, and it is not the dangerous one. The dangerous version runs with no feeling at all. It is built into the very judgments we spent last week admiring. We did not devise those judgments; we inherited them. And an inherited sense of who is credible arrives with its settings already in place: this kind of person is sober, that kind is excitable; this voice is authoritative, that one is probably overstating things. Those settings do their work silently. The doctor who waved me off did not dislike me; she very likely liked me well enough. The discounting happened underneath all that, in the part of her judgment that had already judged my type before I had finished speaking.
Consider what that did. It lowered my credibility below what my reliability warranted. That is the precise shape of it. I was not, in fact, an unreliable reporter of my own body that day: there really was something there, and I had described it accurately. My reliability was fine. What was marked down was something else, something attached to how I came across, and it lowered the credibility I was granted beneath what the facts warranted. The judgment came out low on a true report — not because the report was weak, but because of who was giving it.
Now carry that shape back to the woman behind the wall, whom I described at the very start of all this. For two years she said the building was making her ill, and for two years she was filed as the anxious one, and the mould was behind her wall the whole time. Because the wall was eventually opened — because we now know she was right — we can run Fricker's test on the case and get a clean answer. Was her credibility lowered below what her reliability warranted? Yes, exactly that. She was right. She had been right for two years. And what lowered the weight given to her word was not a record of her being wrong, because she had no such record. It was a settled little verdict about the kind of person she was — the worrier, the one always sure something is wrong — sitting inside people's judgment and doing its work with no one deciding anything. By Fricker's definition, that is testimonial injustice, plainly. But notice the single thing that let me say so with any confidence: the wall was opened. A fact came in from outside and settled it. Hold on to that, because it will matter more than anything else tonight.
So that is what Fricker puts on the table, at full strength. A wrong that is not rudeness and is not cruelty and often is not even intended — a discounting of a person's word that tracks who they are rather than how reliable they are. A real wrong, in the one place where it harms a knowing creature most: it tells them that their word about their own life does not count. Now I have to take it and test it against the view I built last week, and see what it does to that view.
Last week I set that view out and told you to be proud of it. Our sense of who is credible, I said, tracks reliability: it grants more weight to the sources that tend to get things right and less to the ones that tend to get things wrong, and that is not prejudice, that is a track record doing its proper work. We believe the doctor over the forum because the doctor tends to be right. We discount the confident uncle because the confident uncle tends to be wrong. The whole defence was that our judgment tracks reliability, and that tracking reliability is the one honest thing such a judgment can do.
Now set testimonial injustice beside that. Because testimonial injustice is, by definition, a low judgment that is not tracking reliability. That is its entire content. The weight given to the woman behind the wall dropped — and, by how it felt, the weight given to me in that office dropped — not because we were unreliable sources, but because of who we seemed to be. When that happens, the low judgment forms in response not to accuracy but to identity. So the standpoint thinker says: some of the low judgments we hand out are not readings of reliability at all. They are readings of who is speaking, presented as if they were readings of reliability. The same low verdict, with something entirely different being measured underneath — and, from the inside, no way to tell which of the two we are looking at.
And here is why this is worse than simply "our judgment is sometimes wrong." Of course it is sometimes wrong; every method of judging has some error, and the veritist grants that freely. Think of a thermometer, off by a degree now and then. If testimonial injustice were just random error, honest noise, one could shrug it off. But it is not random. It is patterned. It runs in the same direction every time. It marks down the same kinds of people — the same accents, the same jobs, the same bodies — not occasionally but systematically, because the bias sits in our judgment as a fixed setting rather than occurring at random. Consider the thermometer again. One with random error still tracks the temperature on average. One that reads low every time a certain kind of person is in the room is no longer tracking the temperature in those rooms; it is tracking the person. And it will report that low reading with exactly the same confidence it reports a true one.
Which brings back the sharpest finding from the first week and takes it one step further. I told you then that a mostly-right judgment hides its own mistakes — that when I filed that woman, it felt identical to the thousand times I had correctly judged someone to be overstating. There is no second signal for the wrong ones. Now you can see why there is none, and why it is not a fault that could be engineered out. The prejudiced low judgment and the honest low judgment come out in the same form, with the same sense of having simply seen who is credible. From the inside, "I am discounting her because she is unreliable" and "I am discounting her because of who she is" feel exactly the same. Both present themselves as accuracy. And this is not a defect in some people's judgment that better people have corrected. It is a limit of the form itself — and, note, a limit that cuts both ways: it is precisely why we cannot be sure a low judgment is prejudiced, and precisely why we cannot be sure it is not.
So consider where that leaves the view I was so proud of last week. It has not been destroyed, and I will not pretend it has. It still works most of the time; all of that was true. But I have shown that it issues a low judgment it cannot itself tell apart from a true one, that reads the same way every time it goes wrong in this particular direction, and that goes low on a truthful speaker exactly when that speaker most needs to be believed. The doctor's judgment read low on me. The whole question is whether it was reading my reliability or reading my type — and from where she sat, those two looked like one thing.
Here is the part I will not soften, because it is the part that does the work. One might want to say: very well, I shall simply stay alert to it, watch for the prejudiced judgments and correct them. And alertness is better than none. But notice what actually settled the one case whose ending we know. The woman behind the wall was not believed because someone in that building stayed alert. She was believed when they opened the wall and the mould was there in front of them — when a fact came in from outside and settled it for good. And that is the practical point, not a grand one: remove that piece of outside confirmation, stand where she stood in year one with the wall still closed, and ask how anyone there was meant to know which kind of low judgment she was receiving. They could not. Not for want of trying. Because in year one, before the wall was opened, the two judgments looked like the same judgment — and most of the time, no wall is ever opened.
That is the real difficulty for the view, and it is real; I will not pretend otherwise, because the standpoint thinkers are right that it is there and right that it matters. But before we conclude that the view is therefore worthless — before we reach for that — the veritist has a question waiting. And it is a harder question than it looks.
So testimonial injustice is on the table; it is real; it is a genuine wrong; and the standpoint thinkers are right to have put it there. The natural next move — the move the internet would have you make — is to carry it back to last week and use it to throw out the calibration view. There: your judgment is prejudiced, the whole thing is corrupt, stop trusting it. Do not make that move yet, because the veritist is about to ask one question, and it is the hardest in the episode, and he can ask it without giving up a single thing he said last week.
Here is his question. He says: I grant you all of it. Testimonial injustice is real, it is a wrong, I am not the man who denies it happens. Now take me to the woman behind the wall, in year one, before anyone opened the wall, and show me from there that her low credibility was tracking prejudice and not tracking a real difference in reliability. Because recall what testimonial injustice means. It means a credibility lowered below what her reliability warranted. And to know it was below what her reliability warranted, you must already know how reliable she in fact was. You must already know she was right. And in year one — wall closed, mould undiscovered — you did not. No one did. So tell me: standing in year one, with what you actually had, which judgment were you making? The unjust one, or the accurate one?
And the honest answer, which I will not dress up, is that you could not tell. Not "it was difficult." You could not. And here is what gives it its force: it is Fricker's own concession, the one I planted last week, returning with real weight. She is the one who insists that not every low credibility is an injustice. Sometimes we mark a source down and are simply right to, because they really are less reliable, and there is no wrong in it anywhere. So now there are two judgments that come out looking identical — the unjust deflation and the justified low mark — and the philosopher who named the first is also the one who tells us the second is real and perfectly proper. They feel the same from the inside. They come out in the same form. And the only thing that ever told them apart, in the woman's case, was the wall being opened: a fact arriving from outside and settling it.
Take that away — and most of the time, in real life, there is no wall that is ever opened, no mould that anyone ever finds, no later moment when an outside fact arrives and settles which judgment you were making — and the two are, as far as anyone can ever know, the same. That is not to say prejudice is not occurring. Plainly it is. It is to say that we cannot, from inside our own judgment, reliably single out a given low reading and declare that one the wrong kind. Because the judgment we would point with is the very judgment in question.
Now — and here I have to be just as hard in the other direction, because the cynic wants only half of this — none of that makes testimonial injustice go away. The veritist has shown that you cannot prove the single instance. He has not shown that there is no pattern. Return to what we found a few minutes ago: the discounting is not random, it runs the same way every time, it marks down the same kinds of people across thousands of rooms. You may not be able to prove that this particular doctor wronged this particular patient — that one case may genuinely be the justified kind. But when the same kind of patient receives the low judgment across an entire system, year upon year, "it was justified every single time" loses its credibility long before "it was prejudice every single time" does. The pattern can be real and visible even where no single instance can be proved. So the standpoint thinker need not win any one case to be onto something true about the whole.
So consider where that leaves the two of them, because this is the actual state of play and I will not tidy it into a verdict. The veritist stands, undefeated: you cannot read prejudice off a single low credibility from the inside; the category of the justified low judgment is real; and folding every discounting into "injustice" would wreck our ability ever to call anyone unreliable, including the people who genuinely are. And the standpoint thinker stands too, undefeated: the discounting that tracks identity rather than reliability is real, it is a genuine wrong, and its pattern is visible at scale even where the instance hides. Neither is the villain. Neither has refuted the other. What we are left with, at the end of tonight, is not an answer. It is a question we can no longer un-ask and cannot yet settle: of the low judgments our sense of credibility is handing us, right now, today, which kind are they?
I am not going to give you the woman behind the wall again to close this out. You have had her, and she is too easy now: you already know how her story ends — the wall open, the mould there, the verdict in. The whole trouble tonight is the cases where the verdict never comes in. So let me give you one of those instead, and ask you to actually carry it out, not merely nod at it.
Picture someone you would discount. Not a stranger — someone in your own life, the one who always has a complaint, the one you have quietly filed as a bit much. Perhaps a relative forever certain the doctors have missed something. Perhaps the colleague who finds a problem with every plan. Perhaps the neighbour who is sure the new development will flood the road. You have one. We all keep a short list of people whose word we routinely discount, and we discount it for what feels like the best of reasons: a record, we would say, of fussing over nothing. Now this person comes to you with something. The relative says this new symptom is different. The colleague says this plan has a hole in it. The neighbour says the water is already pooling where it never used to.
And here is the exercise, which is harder than it sounds because your own judgment will resist it. Try to settle it. Right now, before you find out, before any outside fact arrives — try to establish, to your own satisfaction, that the low mark you hold on this person is tracking how reliable they really are, and not tracking the fact that you find them tiring. Go on. Reach for the evidence, and watch what you reach for. You reach for the times they were wrong before — but those reached you through the same judgment that had already decided what kind of person they were, so they are not clean. You reach for how they come across — but how they come across is the very thing testimonial injustice runs on. You reach for your own sense that you are being fair — but a prejudiced low judgment and an accurate one come across the same way, so that sense does not sort them either. Everything you reach for is already coloured by the very judgment you are trying to check. And the one thing that would settle it cleanly — finding out for certain, the way the opened wall settled it for the woman two years too late — is not available here. Not today, and for most of the people on your list, most likely not ever. The low judgment you hold on this person, the one in front of you now, is either the justified kind or the unjust kind; and short of that outside confirmation, the two look like one thing, and you cannot tell which it is.
That is the exercise: to be left exactly there, with no outside fact arriving. That is the whole of what tonight was for.
Now let me say what tonight does not license, because there are two easy answers waiting here, and both of them duck the difficulty. The first says: well then, believe them — believe the discounted, side with the fussy ones, give everyone's word full weight again. No. We ruled that out in the first week, and it stays ruled out, and now you can see exactly why: the relative certain the doctors missed something is sometimes merely frightened, the colleague is sometimes merely negative, the neighbour is sometimes simply wrong about the water. Reversing the judgment is not seeing clearly. The unjust low reading and the justified one look the same from the inside, which means "just believe them" is as blind as "just doubt them"; you would only be guessing in the other direction.
The second answer is quieter, and more tempting, and I suspect it is the one some of you are reaching for now to make the evening sit more easily. It says: very well, but that is the rare case, one in a hundred; my judgment is basically sound; the woman behind the wall was the exception; I can file this away as a sad story and go back to trusting my own readings. I cannot let you have that one either, because it is exactly the comfort the whole episode was built to remove. The point was never that our judgment fails one time in a hundred. The point is that we cannot tell which one in a hundred it is failing on — every one of them feels like the ninety-nine. You do not get to file tonight under "rare exception," because the whole trouble is that the exceptions do not announce themselves. They look just like the rule.
So here is what you leave with tonight. Last week you left proud of a fair, well-calibrated judgment aimed at the truth. You still have it, and it still mostly works; I have not taken that from you and I will not. But you now know one further thing about it, and it is a thing you cannot un-know. It issues a low judgment that it cannot, from the inside, tell apart from a true one — a low mark that may be tracking who a person is rather than how reliable they are — and the only thing that ever sorts the two is a fact arriving from outside, which most of the time never arrives. You came in able to feel the difference between fair discounting and unfair discounting. You leave knowing that the feeling was never the thing that could tell them apart.
This is the third of nine, and we have tested the view only once. Next week the difficulty grows stranger, because so far at least the person had words: the woman could say the building was making her ill, even if no one took it. Next week we turn to the people who are wronged before they can speak at all — those for whom our shared language offers no way to say what has happened to them. And to the claim that turns the whole thing over: that where you stand, the very position supposed to bias you, may let you see more than the people at the centre, not less.
But that is next week. Tonight, just this. The next time you give someone's word less weight — and you will do it within the hour; we all do — you will not be able to tell, in the moment, whether you are hearing how reliable they are or hearing who they are. And the thing that would settle it for certain — an outside fact, the truth arriving from somewhere beyond your own judgment — rarely comes in time, if it comes at all. Carry the question anyway.
Thanks for listening. I'll see you next time.