A high-school junior opens her laptop on a Tuesday night and types a question into a chatbot. She is writing a paper on the civil rights movement. The model returns three confident paragraphs, well-organized, lightly cited, calm in tone. She copies the structure, rewrites it in her own words, and turns it in. Her teacher tells her it is one of the better papers in the class.
Nothing in that exchange feels like power. It feels like help. That is the part this episode wants to slow down on.
Where this episode sits in the season
For six episodes, this season has traced a single relationship — knowledge and power — through institutions, laws, workforce, restriction, and the daily life of a child inside a school. Last week, in The Hidden Curriculum, the argument turned inside the student, where the architecture of knowledge becomes a kindergarten teacher’s read on which six-year-old is curious and which one is a problem.
This week, the gatekeeper changes shape again. It is not a state legislature. It is not a textbook adoption committee. It is not a teacher’s instinct in the moment. It is a model. And the model is now sitting between a great many learners and the information they are looking for.
What the episode argues
The argument is not that the technology is evil. The argument is more careful, and harder.
A large language model is a record of what was written down, digitized, scraped, and weighted heavily enough to survive the training run. That record is not the world. It is the dominant interpretation of the world — the books that got published, the scholarship that got indexed, the journalism that got archived, the language that got translated into English. The mini-narratives the show has been threading all season — community knowledge, oral tradition, scholarship from outside the U.S. and U.K. academy, the lived sense-making of people who were never asked to write a chapter — those are present in the record only thinly, and sometimes not at all.
When a model trained on that record speaks back to a learner with the cadence of a confident teacher, it is doing something the show has a name for. It is collective knowledge construction with most of the collective missing. It is banking education at industrial scale.
What you’ll hear next week
Three threads, one episode.
The first is the corpus question — what was actually in the training data, and what wasn’t, and why “the internet” is a poor proxy for human knowledge.
The second is the confidence question — why a calm, well-organized answer feels true, and how that feeling does work the model itself cannot do.
The third is the pathway question — how this changes the route between a learner and the knowledge that would actually meet her, and what educators, learners, and communities can do about it now, not in a future version of the technology that has not yet shipped.
We will close with a “do this this week” that is, for once, the same for all three audiences.
The landing question
The hidden curriculum taught a child, every morning for twelve years, who the school was for. The new question is whether the model, trained on a record that was never the whole record, will teach a generation of learners the same thing — only faster, more politely, and with footnotes.
That is the conversation next week.
The Cultural Context of Knowledge, Season 2, Episode 7: “AI as the New Gatekeeper: Whose Knowledge the Model Was Built to See” — drops next week. Listen on Spotify, Apple Podcasts, or Amazon Music.
Catch up on Season 2 Episode 6, “The Hidden Curriculum: When Researchers Chose the Word Violence,” wherever you get your podcasts.
DEB