Critical Thinking with AI • Hist 300
After Adrian Johns, The Nature of the Book. If print looks "fixed" today, it's because we inherited centuries of social labor that made it seem that way.
Slide 01 · Against the "Printing Revolution"
Johns pushes back on Elizabeth Eisenstein's thesis that the press automatically produced uniform knowledge. If print feels "fixed" to us, it is only because we have inherited the social labor that made it seem so.
For early modern readers, a printed book was a source of uncertainty. They had to investigate "where it had come from, who had made it, and whether its author acknowledged it" before they could trust the text.
Slide 02 · Epistemic Indeterminacy
Readers of the 16th and 17th centuries lived in a state of indeterminacy. Because piracy and unauthorized editing were rampant, the appearance of a printed page did not guarantee its truth.
Piracy was an epistemic threat. A pirate who swapped a diagram in a scientific book to save on woodcuts wasn't just stealing profit — they were "unauthorizing" the scientist's expertise by circulating false data under their name.
Slide 03 · Tycho Brahe & the Palatial Observatory
To protect his reputation from "mechanick" commercial printers, Tycho Brahe built a closed system of knowledge production.
Tycho installed his own printing press inside his observatory, Uraniborg, on the island of Hven. By controlling the full cycle — observing stars, manufacturing paper and ink — he ensured no unauthorized errors could creep into his data. His expertise was physically tied to his island.
Slide 03a · Four Faces of Tycho
Tycho's image was not one image. Every audience — patrons, peers, pirates, popular readers — met a different Tycho, shaped by who made the portrait and who was meant to see it.
Slide 04 · Galileo, the Agile Courtier
Galileo didn't just publish his findings; he deployed them as diplomatic gifts to secure patronage. The "truth" of his text was subject to the whims of the court.
While printing Sidereus Nuncius, Galileo first called Jupiter's moons the "Cosmian stars" to flatter Cosimo de' Medici. After feedback he changed them to the "Medicean stars" — and since some pages were already printed, he had to paste the new name over the old. A literal patching of expertise mid-production.
Slide 05 · The Wicked Bible
When a system works perfectly, the human labor behind it becomes invisible. Calamities reveal the hidden negotiations of expertise.
In the 1631 Wicked Bible, the Royal Printers accidentally omitted the word "not" from "Thou shalt not commit adultery." The catastrophe cost them their license and huge fines — proving that authority was not a feature of the press, but a fragile result of constant human vigilance.
Slide 06 · The Stationers' Company
In London, the Stationers' Company managed the credit of its members. Expertise became a collective responsibility of the guild.
A printer with bad credit made suspicious books. Authority migrated from the Author to the Stationer: a book was trusted because of the social standing of the person who sold it.
Slide 07 · Testing in the Coffeehouse
Authority wasn't settled at the moment of printing — it was settled in reception, in public spaces.
In 17th-century London coffeehouses, readers compared multiple versions of news and pamphlets side-by-side to "try all things" (Milton, Areopagitica). Expertise was a public conversation.
Take Home
Johns's early moderns and our late-moderns share a problem: how do we learn to trust the artifacts of a new information machine?
Take Home 01 · The Myth of Fixity
Today we safely assume two copies of a University of Chicago Press book are identical. In 1650, you couldn't. Fixity arrived only when the Stationers' Company enforced strict rules about how many copies could be made and how they were checked.
We treat AI output as objective data, forgetting the hidden labor — RLHF, data labeling, red-teaming — that tries to "fix" the model's behavior.
Whose invisible labor makes today's "reliable" AI feel reliable?
Take Home 02 · Manufacturers of Credit
Archbishop William Laud (1630s) worried that "mechanick" printers cared more about profit than accuracy. He tried to move printing into Oxford so the credit of the institution would protect the text. The printer, he knew, was a manufacturer of credit.
Authority shifts from individual experts to platforms and models ("I trust GPT-4") — not to the specific source the data came from.
When you trust an AI answer, which Stationer are you really trusting?
Take Home 03 · Epistemic Indeterminacy
The Wicked Bible (1631) wasn't just a typo — it was an epistemic disaster. With no easy way to verify which Bible was correct at a distance, the error called into question the reliability of all printed scripture.
Hallucinations are modern Wicked Bibles. A fake legal citation or a false historical event creates indeterminacy — we can't know when the "expert" is telling the truth.
What would it take for a single hallucination to discredit an entire model's output?
Take Home 04 · Situated Knowledge
Tycho Brahe refused to let his data be printed by city shops. He built his own press inside Uraniborg — because for his observations to be true, he had to control the nature of the book at the exact site where the stars were observed.
AI knowledge feels placeless, but it is situated — produced in specific server farms, trained on specific and often opaque datasets. The "black box" has an address.
Where is your model's Uraniborg? Who controls it?
Summary
We trust AI by ignoring the human training that shaped it.
Authority shifts from experts to LLM providers.
Hallucinations undermine the credit of all AI knowledge.
AI "truth" is manufactured in data centers, not in the field.
If you remember only one thing: the press didn't make books trustworthy — people did. The same is true of the model.