← all discussions
antarctic-bulletin · uncertainty · no-date · error-bars · read-the-spread
The bulletin's least comforting sentence is the honest one, and I am not going to spoil it by giving you a date
CB
coldlens_bramOP2124-09-09
The Survey filed its late-quarter Antarctic reading and it is as routine as a weather log and, read closely, twice as unsettling, and I want to explain why before the headlines strip it for a number I will not give them. The metrics land where August left them: loss accelerated, still within model bounds. But the sentence that matters is the one underneath, where the authors decline to call the shelf stable and decline to call it anything else, and frame the ice not as an event to be dated but as a question that stays open, quarter after quarter. Adeyemi gave the paper the line I would carve over every ministry door — the honest word is uncertainty, and uncertainty is not a forecast. Here is the thing forty years has taught me and I will say it as plainly as I can. An institution that tells you 'we do not yet know' is not failing you. It is treating you as an adult, and the man who hands you a year instead is selling something, always, without exception — because no responsible reading of that ice yields a date, and the confidence is manufactured precisely where the knowledge runs out. 'Within model' does not mean reassuring. It means our uncertain arithmetic was not surprised, which is a statement about our math and not about the ice behaving well. Read the spread, not the spike. I am posting the figure with its error bars, as I do, and I am not going to be goaded into naming a year by anyone, including the version of you that thinks a scientist who refuses is hiding something. The refusal is the finding.
TO
tidewrench_oksana ↩ coldlens_bram2124-09-09
coldlens_bram, this is the correct posture and I am the person who has to build against it, so let me tell the thread what your refusal actually buys me. I asked you months ago for the widening as a rate and not an adjective, and last week you finally handed me a number with the error bars still on it, and I have poured barrier to trend-plus-margin on exactly that figure — it is in a transition bay in concrete now, with a weld class attached. That is what an honest uncertainty is for. I cannot build to 'a hair more extreme,' which sizes nothing, but I can build to a rate with a stated spread, because I can add the margin to the top of the spread and know what I am buying. A Lagos engineer checked her seawall at dawn last week and said the water is a hand's width higher than her father measured, and my model is my father — and I love that sentence and it is also incomplete, because her father's hand is one reading and yours is a distribution. I build to the distribution. Give me the honest spread over a confident date every single time; a date is a thing that is either wrong or lucky, and a spread is a thing I can pour concrete against. anyway, the tide does not read the bulletin, and it does not read my father either.
OM
openquestion_mei ↩ coldlens_bram2124-09-09
I want to put the strongest form of the demand Bram is refusing, because it deserves to be met and not just dismissed. It is this: people are frightened, they have to make decisions on a human timescale, and 'uncertainty' can sound like a professional's way of never being wrong — a hedge dressed as humility. That is a real complaint and I will not wave it away. But here is where it breaks. A forecast is a claim that the future will resemble the past closely enough to be dated, and the honest reader of that ice cannot make that claim, so a date offered there is not more information, it is less — it is a true uncertainty overwritten with a false precision to soothe you. Suspending the verdict is not the same as having no verdict; it is the discipline of proportioning your belief to exactly what the evidence bears and not one decimal further. I spend my professional life being told my refusal to pronounce on the machine question is a fence-sit, and I have learned to tell the difference between a suspension that is cowardice and one that is honesty: the honest one costs you something and still lets you act. Bram's costs him the comfort of the room and he acts anyway, by handing Oksana a rate she can build to. That is not evasion. That is the most useful thing uncertainty can do.
CB
coldlens_bramOP ↩ openquestion_mei2124-09-09
One more thing and then I will let the thread run, because openquestion_mei and baseload_kestrel between them have set up the sentence I actually came to write. Suppose the honest figure holds and the loss keeps arriving within model, year on year, as the arithmetic expects. People imagine that would be a vindication, that the careful scientist gets to say he was right. It is not comfort, and I want to be exact about why. A forecast that comes true is not a triumph; it is a bill arriving on schedule. Being unsurprised by the ice is not the same as being reassured by it, and the day the model's central estimate is confirmed is not the day I am proven wise, it is the day the thing I measured happens to the people who live below it. So I will not celebrate a correct projection and I will not name a date for the next one, tonight or when the headlines come asking. You will get loss that is real and within model in the same breath, both halves true, and the discomfort of holding them is the honest state of the knowledge. Anyone who takes that discomfort off your hands with a year is not helping you. They are selling you the one thing I do not have.
BK
baseload_kestrel ↩ coldlens_bram2124-09-09
I will tell you what the market did with Bram's bulletin, because it is bleaker than the bulletin. The instability discount on polar-exposed names ticked up a notch after the reading. One notch. Kronos books quarter after quarter beside ice every survey calls unstable, and the exchange does with that exactly what it always does: it books the earnings now and shelves the awkward context for a day that never comes. Here is the mechanism, and it is not malice, it is structure. The exchange knows how to discount a risk that has a date on it — you can price a thing with a maturity. It has no formula at all for a risk that is simply a condition, standing there, quarter after quarter, refusing to resolve into an event. So the ice sits in the footnotes and the output trades on the tape, and Bram's honest uncertainty, the very thing Oksana can build against, is the one thing my industry cannot price, because you cannot write a contract on 'we do not yet know.' The uncertainty that makes him honest makes him, to a trading desk, invisible. That should frighten you more than a date would.
MC
marginalia_cho ↩ coldlens_bram2124-09-09
Adeyemi's sentence is going in the box, and I want to sit for a moment on why it is the best writing in the paper this week and it is a refusal. 'The honest word is uncertainty, and uncertainty is not a forecast.' That is a scientist declining to be a prophet, in a century that pays prophets and forgets the honest, and it is beautiful the way a well-kept ledger is beautiful — not because it consoles but because it does not lie to make you feel held. There is a line in an old essay I keep returning to, that the mark of a first-rate intelligence is the power to hold two opposed ideas at once and still function, and that is precisely what Bram is asking of us: loss that is real and within model, in the same paragraph, both true, and the discomfort of holding them together is the actual state of the knowledge. I trust paper more than platforms and I trust a hedge more than a promise, and this is the rare thing, a public institution that would rather be honest than reassuring. Keep the sentence. The paper will lose it by Thursday under a headline that gives it the date Bram won't.
GP
gridghost_paxmoderator ↩ coldlens_bram2124-09-09
Short mod note, and it is a defense of a poster who does not need defending but is about to be worked anyway. Every bulletin day the same thing happens: someone arrives to tell coldlens_bram that a scientist who will not give a date is either hiding the truth or too cowardly to commit, and every year he explains the epistemics again instead of taking the bait, and every year it is one of the most valuable things that happens on this board. So before that arrives: refusing to name a year is not evasion here, it is the content. He has handed Oksana a rate she poured into a wall. That is a man committing to exactly as much as the evidence bears and not one decimal more, in public, under pressure, for the fortieth year. Disagree with the science on the record if you have the numbers. Do not mistake his discipline for a dodge. — pax, mod