Doctoress. Hostess. Duchess.
Notice what they have in common. Each word defines a woman either in relation to a man or as a smaller version of a man’s craft.
A duchess is a duke’s wife. A hostess is a host’s helper. The relational form is the form that survives.
Seamstress and tailor are the same trade. Housekeeper and butler. Cook and chef. The female version is the one we keep. The male version is the one that signals seniority.
That is misclassification at the level of the dictionary. Before any chronicle is written. Before any patent is filed.
The category is in the language
Mr and Mrs Adam Johnson. The wife has no name.
This is not the past. Wedding invitations still print like this in 2026. The woman is included by being absent. Her existence is granted by being denied.
Earlier posts in this series traced the encoding mechanism through the corpus that the model is trained on. Selection, weighting, output. The pipeline. The pipeline does not begin in a manuscript. It begins in the words available to name what people do.
If “seamstress” is the only category available for a woman who makes clothes, the historical record will file the woman who makes clothes as a seamstress. Not as a tailor running her own shop with apprentices. As a seamstress. The data was correct. The category was wrong.
Women are not absent from the historical record. They are filed under the wrong category.
Filed in the household
Behind every great man is a woman.
The phrase is true. It alludes to the household. The household ran on women’s labour while the man had time to think. We accept the line because it grants women a soft credit and asks no further questions.
The further questions are the ones that matter.
Who gathered the data? Who actually had the ideas? Who did the work?
John Stuart Mill wrote in his own autobiography that Harriet Taylor was the co-author of his major work. His own claim, in his own voice. Most readers of On Liberty and The Subjection of Women still cannot name her. However, he didn’t add her as an author. Whose decision was that?
I had to look up Harriet Taylor Mill’s name when I started writing this post. Meaning, I’m inside the same system I’m naming. The encoding fault runs on me too.
Kant. Nietzsche. Pick any name in the Western canon of philosophy from the eighteenth century onward, and the same household sits behind it, run by women whose names entered no record because the category for what they did was “support”.
The Harvard Observatory in the late nineteenth century employed dozens of women to compute the positions and spectra of stars. Henrietta Leavitt worked out the period-luminosity relationship that became the cornerstone for measuring distance across the universe. Annie Jump Cannon designed the stellar classification system that astronomers still use. Williamina Fleming catalogued more than ten thousand stars. The men who supervised them are remembered as astronomers. The women were “computers”. Maria Popova’s Figuring documents the whole operation.
The category was “computer”. The work was groundbreaking science.
Filed under celebrity
Hedy Lamarr and George Antheil filed a patent in 1942 for frequency-hopping spread spectrum, the underlying mechanism behind the radio technology that now runs WiFi, Bluetooth, and GPS.
US Patent 2,292,387. Granted in her name and Antheil’s. The US Navy was offered the design that year.
The National Inventors Council declined to develop it. The Council’s response, as recorded in Science in 2018 (DOI 10.1126/science.aar4304), was that Lamarr could better serve the country by using her status as a celebrity to sell war bonds.
She sold war bonds. The patent expired in 1959 without being used. The technology was independently rediscovered, militarised in the 1960s, and commercialised in the 1990s. Lamarr was formally recognised for the invention in 1997 with the Electronic Frontier Foundation Pioneer Award. She was in her eighties.
The category available for her was “actress, beautiful”. The category required to read her patent on its merits was “inventor”. The Council had access to both. It chose the first.
This is the encoding fault in real time, in 1942, with the response printed in the record. The data was correct. The patent worked. The category that determined whether the work would be developed was the category the work could be filed under, and the woman who filed it had a category attached to her already.
Filed under consort
Earlier posts in this series documented four mechanisms of suppression in the historical record of women who ruled or governed. The previous post unpacked one of them: the chronicle gap, where medieval narrative histories filed the activity of women regents under “holding the household together” while administrative records preserved the actual decisions.
Three more sit beside it.
Title erasure. Margrete I of Denmark controlled three kingdoms through the Kalmar Union for decades, longer than most kings of her era ruled one. She was never formally Queen Regnant. Irene of Athens ruled the Byzantine Empire from 797 to 802 and declared herself Basileus, the male form of emperor, because she understood that the female form meant the emperor’s wife. She fought the vocabulary because the vocabulary was the suppression.
Credit to the advisor. Catherine de’ Medici governed France as regent for nearly thirty years. For centuries, historians credited the men around her. Anne of Austria’s regency was attributed to Cardinal Mazarin’s genius, despite the fact that she had outmanoeuvred her dead husband’s will to take the regency in the first place.
The husband as cover. These women ruled territories, negotiated treaties, and commanded armies whose names do not appear in the standard histories because their husbands were technically alive, formally credited, or nominally present. The activity was theirs. The historical category was his.
Four mechanisms. The activity is recorded. The framing strips the agency. The category fits the framework, not the actor.
Filed under amateur
The same move ran in painting, in invention, in scholarship, all at once.
Women were legally barred from attending life drawing classes in most European academies until the late nineteenth century. The justification was that male nudes were unsuitable for female eyes. Without that training, they could not paint the large-scale history paintings the academies treated as serious art. They were confined to portraiture, still life, and domestic scenes. Those genres were then ranked below history painting. The exclusion produced the limitation, and the limitation was then used as evidence of lesser ability.
Artemisia Gentileschi painted some of the most powerful works of the Baroque period. She was filed for four centuries as “the daughter of Orazio Gentileschi”. Clara Peeters was a founder of still life as a genre. The Prado gave her the first solo retrospective of any woman painter in its collection. In 2016.
Women were barred from European universities until the late nineteenth century. Oxford did not grant women full degrees until 1920. Cambridge held out until 1948.
Until 1948. One human lifetime ago.
The recovery scholarship that names what was lost is, on the historian’s clock, brand new. Women’s history as a recognised academic discipline is roughly fifty years old. That is one career.
The category is the instrument
Hedy Lamarr in 1942.
A woman in 2026, asking an AI for advice on how to present her own work, was told to open with “Maybe, I’m not right about this.”
Two moments, eighty-four years apart, and the mechanism is identical.
The category attached to the person determines whether the work is evaluated on its merits. In 1942, the category was “actress, beautiful,” and the patent sat unused. In 2026, the category is applied by a system trained on the historical record, and the advice it returns is shaped by what that record encodes about women in professional situations.
The Rational Disengagement post in this series documented the logical endpoint: the only way to receive advice at the correct professional level was to remove the categorisation entirely. To ask for the advice that the system would give a man.
That is not a personal workaround. That is the historical suppression mechanism running in real time, on a new substrate.
Women are not absent from the record. They are filed under the wrong category. The category was the suppression. AI is now being trained on that record, and the filing system comes with it.
Leave a Reply