The Evidence That Followed

TAM-EAR.3-05 · The Standard with Ears · The Approximate Mind

Priya is a clinical dietitian in a teaching hospital, nineteen years in, and she keeps her patient food diaries. Not the data extracted from them, the diaries themselves: cheap notebooks, printed forms, a shoebox and then several shoeboxes of what people actually wrote down about what they actually ate, in their own hands, with their own spellings. She keeps them because the diaries taught her the one thing her training undersold, which is that nobody eats a diet. People eat compromises, in kitchens, under pressure, and the distance between what a protocol prescribes and what a Tuesday allows is where all of her real work happens.

Three or four times a year, for her whole career, a wave has come through her clinic. A fat that is suddenly poison, a fat that is suddenly medicine, a seed oil, a fermented thing, a fasting window, a powder. She can feel a wave arrive before she can name it, because her patients start asking the same question in the same words in the same week. The waves used to build over months, from a book or a talk show, and she could watch one coming. The recent ones arrive in days. And in the last two years something has changed in the question itself. Patients used to ask her whether the thing was true. Now they arrive having asked, and the system they asked has answered fluently, with a summary of the enthusiasm, sometimes with the caveats attached and sometimes with the caveats smoothed into the confidence, and what they want from Priya is no longer information. It is permission to disbelieve something that sounded settled.

The pace of finding out

#

Nutrition science is slow for reasons that are not fixable, and stating them plainly is the ground this essay stands on.

Feeding is the hardest exposure in epidemiology. People misreport what they eat, everyone eats many things at once, the outcomes of interest take decades to arrive, and the randomized trial that would settle a question cleanly would require locking thousands of people into controlled diets for twenty years, which cannot be done and should not be. So the field advances by triangulation: observational cohorts, short mechanistic studies, the occasional heroic trial, each weak alone, converging slowly or failing to. The honest answer to most nutritional questions is a confidence interval and a decade of waiting, and the field’s history is a record of confident reversals, dietary fat and eggs and margarine, each reversal itself taking years to propagate.

That slowness was survivable under the old information regime because the hype was slow too. A diet craze traveled by book tour and word of mouth; it took seasons to build and seasons to fade, and somewhere in the overlap the slow correction machinery could at least begin to engage it. The two tempos, enthusiasm and verification, were mismatched, but they were mismatched within an order of magnitude.

The counting machinery broke the ratio. Enthusiasm is exactly the material the engagement filter selects for: a food that heals or a food that kills is inherently inflammatory in the technical sense this argument uses, it moves under its own power, and the machinery amplifies it in proportion to the movement. And the monetizable branch arrives fused to it, because every wave has a powder, a plan, a product attached within days, the supplement market being one of the purest converters of belief into revenue that exists. A nutritional claim is the rare cultural object that fires both filters at maximum from the moment of its birth. Verification, meanwhile, fires neither. A twenty-year cohort result converts into no product and inflames no one, least of all when it reports, as it usually does, that the effect is small and conditional.

The claim travels on both filters and the correction travels on neither, and the gap between their speeds is now the main fact about what a population believes it knows about food.

The loop that manufactures its own confirmation

#

What is new is not the speed alone. Fast hype is old. What is new is what the hype becomes when the machinery that amplifies it also feeds the systems that answer questions about it.

Follow one wave through the loop. The enthusiasm is posted, in volume, because the filters reward it. The volume is harvested; it is text, and text is the intake. The systems trained on the intake now hold the wave at depth, because the frequency column keeps what is repeated, and the enthusiasm was repeated a millionfold while the skepticism was repeated a hundredfold. A patient asks the system about the powder, and the system answers from the column, fluently, in proportion, and the proportion is the enthusiasm’s. The answer is posted, discussed, harvested in turn. The next model holds the wave a little deeper.

At no point in the loop did anyone find anything out. The question of whether the powder does anything is exactly as open as it was on the first day, sitting in a trial registry with a completion date four years off. But the epistemic surface, the thing a person touches when they ask, has hardened. The enthusiasm has become the consensus of the corpus, and a corpus consensus is what these systems are built to report. The amplification became the evidence, in the only operational sense of evidence most askers ever encounter: the thing that comes back when you ask, said confidently, from many sources, agreeing with itself.

Priya meets the loop at its output. Her patients are not credulous; they are reasonable people who asked a reasonable question of the most fluent instrument they have ever held, and the instrument reported the state of the column, and the column holds the wave. When she counsels against it, she is one voice, with a shoebox of food diaries and nineteen years, against the summarized agreement of what feels like everyone. The structure of her position is familiar from elsewhere in this corpus: her knowledge is unfalsifiable to the apparatus at the moment it is most needed, and the apparatus is not wrong on its own terms, and its terms are the only ones in the room.

By the time the trial reports, the question has been settled for two years by the volume of the asking.

What follows the evidence

#

The reversals still come. The cohort completes, the trial reports, the effect is small or absent, and the slow machinery of the field grinds out its correction. What this domain shows is what the correction now arrives into.

It arrives into a column that holds the enthusiasm at depth and will hold it for the operational life of every model trained during the wave. It arrives as one document against a million, unamplified, because a null result fires neither filter. And it arrives late in a culture whose waves have already moved twice, so that the correction, when it surfaces at all, attaches to a question nobody is currently asking. The old reversals overwrote the old errors slowly, in the same channels that had carried them. The new errors are fossilized in the training strata of the systems that answer, corrected in a literature the systems summarize less often than they summarize the wave, because the wave is what the harvest weighted.

Some of Priya’s counsel does not survive either regime, and the essay should say so. Clinical dietetics has its own reversed certainties, and she has unlearned protocol she was confident of in 2010. The difference she stands on is not that she is right. It is that her way of knowing can lose, patient by patient, in follow-up numbers she has to sit with, and that losing changes what she says next. The loop has no such moment. Nothing in it is downstream of the powder working.

We wonder whether verification can ever be made loud, or whether loudness is structurally the property of claims and never of their testing, so that every instrument built to summarize what is said will faithfully report the enthusiasm of its era back to it as knowledge.

The shoeboxes are the part of her practice no wave has ever touched. A man writing down, in pencil, that he had the sweet at his grandson’s birthday and is sorry. A woman recording the same breakfast two hundred times, and then, one Tuesday in October, not. Nothing in them converts, and nothing in them inflames, and no system will ever be trained on them, because they are in a cupboard, in handwriting, in the third class.

Priya reads a few pages most weeks, usually before a difficult consultation. Not for data. For calibration, she says, though when pressed she puts it differently: to remember the tempo at which anything about eating is actually true, which is one person, one kitchen, one Tuesday at a time, for years, which is a speed no filter has ever selected for and no wave has ever moved at.

This essay is the fifth of the domain arc of The Standard with Ears and the arc’s clearest case of the engagement filter of The Standard with Ears fused to the monetizable one from birth. The loop it traces, enthusiasm harvested into the column and reported back as the consensus of the corpus, applies the frequency argument of the Midden’s object arc: what is repeated sits deep, and depth is what the answering surface reports. The dietitian’s position, right in a register the apparatus cannot read and answered by an instrument that is correct on its own terms, stands beside the nephrologist of The Counterparty. The unread diaries join the arc’s running third class.