For six weeks we have been looking for a firm basis for a single conviction — that we owe the people who come after us a world worth living in. Each episode reached for a foundation under that conviction, and each foundation failed under examination. A victim who, on inspection, turned out not to be there. An impartial standard too cold for anyone to live by. A claim that no one yet alive was present to make. And last week, the economists' number — the one that promised to do the moral work for us and hand back a settled figure — turned out to be a moral choice presented as arithmetic.
Underneath all of that, week after week, we have been making one quiet assumption. We have assumed that the hard part is finding the basis. That the whole difficulty is not yet knowing what we owe — and that once we knew that, the rest would follow: we would know what to do.
Tonight I want to take that assumption away as well.
Here is the point that is easy to miss. Suppose we had the basis. Suppose that tomorrow morning someone handed us the answer, complete and settled — this is exactly what we owe the future, and this is exactly why. We would still have to act on it. And to act on it, we would have to know things about the future that no one can know. How bad it actually becomes. Who the people downstream turn out to be, and what they have. Whether the thing we give up a great deal to provide is in fact the thing that helps them — or whether we guessed wrong and spent everything on the wrong rescue. The duty could be settled tonight, every word of it, and we would still have to act toward a future we cannot see clearly enough to act on with any confidence.
This is Philosophy for Us — philosophy for everyone, no degree required.
This is the seventh episode of nine. Last time we examined the one way out that promised to remove the moral burden entirely — the economists' number, the price they had quietly placed on the future — and we found a moral choice sitting inside it, presented as arithmetic. There was no escape there. Tonight the question itself changes. Not what do we owe — that has been the whole month, and it remains open. Tonight the question is how we are to act toward the future at all, when we cannot trust a single one of the things we would need to know in order to act well. That turns out to be a distinct problem, with difficulties of its own. And before the end of the episode we will watch the very tool a careful person would reach for — a sober rule for deciding under uncertainty — point first at one disaster and then, turned around, at a different disaster in the opposite direction.
So tonight we are not searching for the basis. We are asking what it is to act well when the facts the decision depends on cannot be known.
Let me make this concrete, because "we cannot know the future" sounds like a truism, and a truism is easy to agree with and then walk past. This is not a truism. It is a specific, structured kind of not-knowing, and you have to see its shape before the rest of the episode makes sense.
Suppose the decision were yours. Not a slogan, but the actual decision. A country is deciding, this year, how hard to push: how much to spend, how much to give up, how quickly to dismantle an established way of living and build a replacement. And suppose you are the one who sets it. You are setting it for the people downstream — that is the whole point, and the one commitment that survived six weeks of argument. So before you set it, you would want to know whether what you are about to impose on your own generation actually arrives, generations later, as a rescue.
And when you reach for the facts you would need, they are not available.
Begin with the simplest. How bad does it actually become? Not roughly — precisely enough to act on. Two degrees of warming, three, four? At three degrees, is it a difficult century that people endure and adapt to, or the century in which some part of the system passes a threshold and does not return? No one can tell you. Not because the work has not been done, but because the honest answer runs out somewhere beyond the reach of anyone's instruments.
Now the next fact, which is harder. Suppose you knew the temperature exactly. You would still not know what it does to the people living through it, because you do not know who they will be or what they will have. Perhaps they are three times richer than we are, and repair in an afternoon, with tools we cannot imagine, the thing that would have set us back a decade. Perhaps they are poorer, and a flood we would absorb is the end of them. You are about to spend your own generation's comfort on their behalf, and you cannot see who they are or count what they have.
And then the hardest. Suppose you knew the temperature, and you knew their wealth. You would still not know what they will value — what counts, for them, as a life worth living; what they would thank you for, and what they would find quaint, or beside the point, or even an insult. You are about to make large sacrifices in the name of giving them something good, and "good" is the term you have the least hold on of the three, because it is theirs to define, not yours, and they are not here to say what it means to them.
Now hold all three together. You are being asked to act — decisively, now, at real cost — toward a place you cannot see, for people you cannot picture, by a standard of "better" you do not get to set. And here is what makes this more than a gap that further inquiry could fill: you cannot study your way out of it. More research sharpens the near term a little and never reaches the far one. There is no expert in the next room with the real figures — no more than there was one, last week, with the true discount rate. This not-knowing is not a hole in what we have found out so far. It is built into the situation. We are acting decades ahead, and nothing from that future is available to us now.
So this is a new kind of problem, different from every problem so far. For six weeks the question was always what is true — is there a victim, is the world worse, does anyone have a claim. Tonight the question is what we are to do when we cannot get at what is true — when the facts the decision needs are not hidden somewhere, waiting to be uncovered, but simply absent, not yet anywhere in the world to be found. That is not a milder problem than the earlier ones. It is a harder one. And it has a definite shape; seeing the shape is the difference between acting under uncertainty and merely flailing under it.
Here is the shape. There are two different things we both call "not knowing," and the whole episode turns on telling them apart, because we reach for the same tool in both cases, and it works in only one.
The first kind we handle confidently every day. Call it the known gamble. You do not know how the coin will land, but you know it is a coin, you know the outcomes are heads or tails, and you know the odds are even. You do not know which card is next off the deck, but you know there are fifty-two and exactly what is among them. An insurance company does not know which house on the street will burn this year, but across ten thousand houses it knows, almost exactly, how many will, and it can set a price that comes out right over time. This is not-knowing with the odds still in hand. And there is a well-developed method for it — the same method the economists were using last week. You take each thing that might happen, weight it by how likely it is, and add the results together; out comes a figure that tells you what the gamble is worth and whether to take it. Weight, multiply, sum. It is the most powerful procedure we have for acting when we do not know the outcome. And where it applies, it genuinely works.
The second kind is the one we are in tonight, and it is not the coin. Call it genuine uncertainty. Here you do not know how it will land, and you do not know the odds either. There is no even chance to rely on, no deck to count, no ten thousand past cases to average — because there is only one future, it has not happened, and it is not quite like anything that has. Ask what the chance is that the century passes the threshold from which there is no return, and the honest answer is not a number at all. It is that no one has a trustworthy one, and anyone who offers a clean percentage is either guessing or treating genuine uncertainty as though it were a coin.
Economists have names for the two — they call the first risk and the second uncertainty — but the plain distinction is the one to keep: a gamble whose odds you know, and an uncertainty whose odds you do not.
Now notice the difference, because the whole hour depends on it. The known gamble and genuine uncertainty look identical from the outside — in both, you do not know what is coming. But to the method they could not be more different. The method runs on odds. Give it the coin's even chance and it works, honestly and usefully. Give it genuine uncertainty and it has nothing to multiply: you would be weighting each outcome by a probability you had made up, adding together numbers you had invented, and then calling the invented total an answer. And here is the damaging part: the method does not fail loudly when you do this. It does something worse. It keeps running. It hands back a result that looks every bit as precise and authoritative as the one it gave you for the coin — except this one rests on nothing, and nothing on the face of it says so.
Notice what has happened, because we met a near relative of this last week. Last week the number ran perfectly well, and the trouble was the choice we found concealed inside it. This week the trouble is one level below that: the number cannot even be produced, because the thing you would have to supply — the odds — is not there to be had. The cost-benefit calculation is not wrong about the far future. It has simply stalled. It has nothing true to work on. And a stalled calculation that goes on producing confident numbers is more dangerous than one that admits it has stopped, because everyone in the room reads the number off the page and acts as though it were still sound.
Which leaves us where the uncertainty left us, except that now we can name the trouble. We cannot run the gamble, because there are no odds to run it on. And we still have to act — the decision does not wait politely for better information. Refusing to choose is itself a choice; the course is set, one way or another, this year, whether or not we put our hand to it. So if the calculation cannot tell us where to set it, what can? That is the question the rest of the episode lives inside. There are rules for acting under genuine uncertainty — rules for what to do when the odds run out — and the one almost everyone reaches for first is the one we have to examine most carefully.
Before we reach for a rule, there is one more feature of the situation to see, because it is the one that turns this from merely uncomfortable into urgent. It concerns what lies at the far end of what we cannot rule out.
Return to the known gamble for a moment — the coin, the deck, the insurer. In an ordinary gamble the wildest outcomes, the ones at the extremes, are so unlikely that they count for almost nothing. In principle a coin could land on its edge and balance there; but that is so close to impossible that no reasonable person planning a bet gives it a second's thought. The extremes are thin. They sit far out at the edge, contributing almost nothing, and you can set them aside and still come out right. That is the world the method was built for — one in which the middle is what matters and the extremes round down to nothing.
Now consider what the economist Martin Weitzman argued about the climate, in a 2009 paper on the economics of catastrophic climate change. Out at the far end of the climate case — the genuinely catastrophic outcome, the one where the system tips and there is no second attempt — the extreme is not thin. It is, in the technical term, fat. Let me say plainly what that means, because it is the whole point. A fat tail means the catastrophe is not as wildly unlikely as the tidy models quietly assume. Not likely — no one is claiming likely. But not the coin on its edge either: a real, stubborn, small-but-not-vanishing chance of the very worst outcome. And the worst outcome is so bad, and so final, that the usual move of setting the extreme aside is not available.
And here is why that single fact changes everything. Run the method on an ordinary gamble, and the terrible outcome, multiplied by its tiny probability, comes out near zero — negligible, as we said. But take an outcome that is unboundedly bad — ruin with no lower limit, which cannot be recovered from or compensated for — and multiply it by a probability that is small but stubbornly not zero, and the result does not come out small. It comes out large enough to overwhelm everything else. That single catastrophic possibility, even at low odds, can weigh so heavily that it dominates every ordinary cost and benefit being weighed against it. Weitzman called this the dismal theorem. And what is dismal about it is not that it says the future is doomed. It is that it says the method itself gives out. The very tool we would use to be reasonable about all this — weigh it, price it, decide — stops producing a usable figure the moment a fat-tailed catastrophe is in the mix, because the calculation keeps tending toward something like "unboundedly bad, at odds no one can fix," and a country cannot be run on that.
Now be careful here, because this result is easy to misread in a particular direction. The dismal theorem is not a proof that we must act, or how much, or how fast. It would be a relief if it were — a clear verdict at last, after six weeks without one. It is not a verdict. It is the opposite of one. It is a proof that the tool meant to hand us the verdict has broken down at exactly the point where we needed it most — on the one question that has the catastrophe in it. So it should not be pocketed as "the mathematics says sacrifice everything." The mathematics says no such thing. The mathematics has given out. That is the news, and it is worse news than a figure predicting disaster would have been — because a figure could at least be argued with.
So put the two halves together, because now the whole problem is in view. We cannot run the gamble, because there are no trustworthy odds. And even where we try to run it anyway, the one outcome we most need to weigh — the irreversible catastrophe — is exactly what makes the weighing impossible, because it is too large and too uncertain to be weighed at all. The calculation has stalled, and the thing that stalled it is the thing that matters most. And still — it bears repeating, because the situation keeps tempting us to forget it — we have to set the course this year. So now, at last, we look for a rule that works when the calculation does not. And we will find one. The difficulty will not be finding it. The difficulty will be living with what it asks.
So here is the rule, and you already half-know it, because it is the one we reach for the moment the calculation fails. It is old, it is sober, and it is not panic. It runs like this. When the worst outcome is unbounded and cannot be undone, you do not sit waiting for odds that will never arrive. You act to keep that outcome off the table — even at real cost, even with no figures to show for it — because some decisions cannot be reversed, and you do not stake what you cannot recover on a probability no one can supply. Guard against the worst case. That is the whole rule. Let me give it its full strength, because it has earned it.
We already live by it where the stakes matter most. You do not let a small child play in traffic on the grounds that you have not seen the actuarial tables and cannot put a clean number on the danger. The possible outcome is so bad and so final that the missing percentage is simply beside the point; you do not allow it, and you do not need the odds to know that. You wear the seatbelt for the crash you will almost certainly never be in, because "almost certainly" is not "certainly," and what lies on the other side of that small gap is your life. That is not carelessness about probability. It is being exactly right about a particular kind of stake — the irreversible one, the one there is no coming back from. Where the worst case is final, the sober response is not to calculate but to refuse. Lay that over the climate's fat tail and it stands up as a serious position: there is a real chance of a catastrophe we cannot recover from, so we keep it off the table, and we do not make the people downstream wait while we argue over the decimal. A careful person hears that and nods. I half-nod myself.
And now — because this is what we do here — we run the objection. Not a weak one. The strongest there is, because to give you the rule without its objection would be to do exactly what I spent last week refusing to do — present something as sound when it is not.
Here it is. "Guard against the worst case" sounds airtight until you ask it the one question it cannot answer: which worst case? Because acting has one too. Pour everything into preventing the climate catastrophe and you have made choices with their own unbounded downsides. You have taken food and medicine from people who are poor right now, whose suffering is not a fat-tailed possibility but a thin-tailed certainty. You may have shut down the very growth and invention the future would have used to save itself. And you have staked the whole present against a catastrophe on which — recall — you cannot put odds, which means you also cannot put odds on its not arriving. The rule said give priority to the bad forecast. But there is a bad forecast whichever way we turn. There is a catastrophe in acting and a catastrophe in failing to act, and "guard against the worst case" was supposed to break that tie — except now there is a worst case on both sides of it, and the rule now recommends opposite actions and cannot say which.
And it grows worse the further you follow it, because the rule has no limit built in. "Prevent the unbearable outcome at any cost, never mind the odds" — followed honestly, all the way, will justify almost anything. Name any catastrophe you can imagine, however far-fetched, and as long as you cannot prove it impossible, the rule instructs you to drop everything and guard against it. That is not careful reasoning under uncertainty. It is being driven by whichever disaster you happen to be picturing that morning. The same rule that looked so sober a moment ago, pushed to its limit, licenses anything at all — and this is a shape you have met before in these episodes: an argument you cannot fault, which nonetheless leads you somewhere you cannot actually live.
So look hard at where we are, and hold both ends of it at once. The calculation failed — no odds to run. We reached, sensibly, for the rule meant to take over when it fails: guard against the worst case. And the rule, looked at squarely, hands us two catastrophes pointing in opposite directions and no way to choose between them — because the one thing that would let us choose, the odds, is exactly what we do not have. The rule for acting under uncertainty turns out to require the very information the uncertainty denies us. I am not going to tell you that guarding the worst case is the answer. And I am not going to tell you it is a fraud and we may relax — because the irreversible catastrophe is real, and the calculation really has broken down, and "relax" is simply guarding the other worst case without admitting that this is what one is doing. Both responses are live. Neither one settles the matter. We are going to have to choose how to treat what we cannot know — and that choice is not a calculation, and never was. It is a stance taken under uncertainty. And we own it, because there is no figure beneath it to take the blame in our place.
Let me set down what tonight actually provided, and then what it cost, because both are real and I do not want them to blur.
The gain first, and it is a tool worth keeping. You came in thinking the only hard part was finding the basis — that once you had settled what you owe, the doing would follow along behind. You leave knowing better: that there is a second hard part, downstream of the first, which does not go away even if we solve the first tomorrow. When you have to act and the odds are not there, you now know what to do — not what to choose, but what to see. You can tell the known gamble from genuine uncertainty, the case that has odds from the one that does not. You know that the calculation does not fail honestly under genuine uncertainty; it goes on producing confident numbers built on nothing, and you know now not to trust the clean face of them. And you know the rule you will reach for first, guard against the worst case, and you know to run its objection before you rely on it, because the worst case has a way of lying on both sides at once. That is portable, and it is yours. It has nothing to do with the climate in particular. It applies to every decision you will make where the stakes are high and the odds are gone.
Now the cost. This was meant to be the way out. Every week we reached for a basis and the basis failed — and tonight we were not even reaching for a basis; we were reaching for a way to act without one, a rule that would let us move with the question still open. And the rule arrived with its limit already showing, like everything else this month. So it does not lift the burden; it adds to it. Seven weeks in, we went looking for a way to act without a settled reason, and the one tool we found gives contradictory guidance. And the basis still has not appeared.
And — I will note it once and move on, because you already know it — the conviction did not move tonight either. We did not even lay a new reason under it this week; we went looking for a way to act and found genuine uncertainty instead. And the conviction is unchanged, exactly as it was in the first episode: it still rules out treating the future as fair game. Except that now we cannot even tell ourselves we will act on it and work out the reason later — because the acting turns out to have its own uncertainty, with its own unargued choice sitting in the middle of it.
One week left before the series is done, and there is one place we have not gone. We have tried every basis for the duty and seen each one fail under pressure. We have tried to act without a settled basis and found a catastrophe waiting whichever way we go. There is one answer left — the one I have kept back deliberately, the deepest version of the question that has run underneath all seven weeks: if not the harm to them, then what? It does not rest on a victim, or a sum, or a contract, or a claim — on any of the things that failed. So it may not fail the way they did. We come to it next, and we come to it last, on purpose.
So here is the one thing to carry out — and do not let it stay a thing I told you. A move like this becomes yours only once you have watched it work somewhere the climate is not bearing down on you.
Set the planet aside. Consider a person — call her anyone. She has a result back from a test, and the result is genuinely new. Not "you have a one-in-ten condition" — there is no one-in-ten, because what has been found in her has barely been seen before and no one has the figures on what it does. It could be nothing at all. It could be the thing that ends her life early, without warning. No odds. Genuine uncertainty. And there is a procedure available — drastic, and irreversible. Have it done, and part of her life is gone for good, whether or not the danger was ever real. Decline it, and if the danger was real she learns that too late to undo anything. Now run the rule on her. Guard against the worst case: the catastrophe is final, so act, take the procedure, do not sit waiting for odds that will never come. A sober response, at least on its face. Then run the objection: the procedure is itself irreversible — she may be destroying a whole healthy life she would otherwise have had, for a danger that never arrives — and "guard against the worst case" cannot tell her which of the two irreversible outcomes is the one to fear, because something unbounded waits through both. Put her in front of the calculation and it stalls; there are no odds to supply it. Give her the rule and it recommends opposite actions and cannot say which. And she still has to decide by Friday. There is no figure anywhere that decides it for her. Whatever she does, she does under genuine uncertainty — and she owns it.
Run that — her, not the planet — and when the calculation fails and the rule gives contradictory guidance, you will know exactly what it is to act toward a future you cannot see. Which is what we have been doing badly with the climate the entire time, and calling it policy.
Thanks for listening. I'll see you next time.