Daniel sent us this one — he's pointing at a study from last year on Irish textbooks that found a disturbing amount of anti-Israel and arguably anti-Semitic material in books being given to twelve-to-sixteen-year-olds. And he's asking the bigger question: who actually writes the history books our children read? Not as a metaphor, but literally. Who are these people, what oversight exists, and where's the line between legitimate curriculum regulation and state propaganda? It's one of those questions that's so obvious once you hear it, you wonder why we don't talk about it constantly.
Because the machinery is deliberately invisible. The phrase "written in the history books" works as a colloquialism precisely because we don't think about the writing part. It just appears, as though textbooks were deposited by storks.
The stork theory of curriculum formation.
Which is actually a pretty good description of how most education systems want you to think about it. But the Irish case blows that apart. In May of last year, the Irish Jewish Museum partnered with the Department of Education — which is itself remarkable — to conduct a study of fourteen commonly used geography and history textbooks for the twelve-to-sixteen age bracket. Twelve of the fourteen contained material that the study's authors described as starkly anti-Israel and arguably anti-Semitic.
Twelve out of fourteen. That's not a few bad apples. That's the orchard.
The specifics matter here. Maps of the Middle East that simply omitted Israel entirely — not disputed borders, not contested territories, just a blank where the country is. Descriptions of the Israeli-Palestinian conflict that used terms like genocide and apartheid without any contextualization of Hamas, without mentioning October seventh at all in the newer editions. One textbook described Zionism as a form of racism.
These are not fringe publications from some activist press. These are the mainstream textbooks approved for use in Irish schools.
Published by Folens and Gill Education, which are the two biggest educational publishers in Ireland. Books that were on the Department of Education's approved list. And the approval process in Ireland is what makes this so instructive. The Department publishes a list of recommended textbooks, but it does not vet content line by line. It's essentially a light-touch model — publishers self-regulate, the Department provides broad curriculum guidelines, and individual schools and teachers choose from the approved list.
The approval is more of a handshake than an audit.
And that's not unique to Ireland. It's actually the most common model globally, especially in English-speaking countries. The assumption is that market competition and professional norms will produce balanced, accurate materials. The Irish study demonstrated that this assumption is, to put it gently, not holding up.
Let's pull back. Daniel's prompt is really three questions folded into one. Who writes these things? How do different countries handle oversight? And the counterpoint — is any state intervention in curriculum just propaganda by another name?
Let's start with who writes them, because the answer surprises most people. It is not, by and large, academics. It is not subject-matter experts. It is not practicing historians or geographers. The textbook authorship industry is dominated by freelance writers, retired teachers, and educational content developers working on contract for publishing companies.
The person writing your child's history textbook might be someone who last studied history formally in university twenty years ago, and who's now writing to a publisher's brief.
The publisher's brief is not "write the most accurate history." The brief is "write a textbook that will get adopted." The incentives are commercial. In Ireland, the biggest buyers are the Catholic school system and schools aligned with the teachers' unions. If the dominant cultural narrative within those institutions leans a certain way on a particular geopolitical issue, publishers have a commercial incentive to reflect that.
Which is not a conspiracy. It's worse than a conspiracy — it's just market logic operating on something that shouldn't be a market good.
That's the phrase. Textbooks carry the authority of neutral knowledge. A news article comes with a byline and a publication that has a known editorial stance. A political speech is explicitly advocacy. But a textbook says "this is what you need to know to pass the exam." That framing makes the ideological content invisible to the student and often to the teacher.
The status of textbook authorship reinforces this. It's low-status, low-pay work. These writers are not celebrated. Their names are not on the spine. In many cases, the "author" listed on a textbook is actually an academic who lent their name and did a light review, while the actual writing was done by an uncredited freelancer. It's ghostwriting for the education sector.
There was a fascinating piece of research a few years ago that looked at the acknowledgments pages of major textbooks and cross-referenced them with the people who actually did the writing. In something like sixty percent of cases, the named author had contributed less than twenty percent of the text.
We have a system where the people writing the books are invisible, underpaid, and working to a commercial brief that reflects the perceived biases of the largest buyers. And then we call the result "the curriculum" as though it descended from a mountain.
The Irish case shows what happens when that system encounters a politically charged topic. The publisher's calculus is: our buyers expect a certain framing of the Israeli-Palestinian conflict. If we deviate from that framing, we risk losing adoptions. If we conform to it, we sell books. No one sat in a room and said "let's indoctrinate Irish children against Israel." The system produced that outcome without anyone intending it.
The banality of textbook bias.
Which brings us to the comparative question. How do different countries handle this? And the spectrum runs from almost no state oversight to complete state control, with every model having its own failure mode.
Let's map it. Where does Ireland sit?
Ireland is on the light-touch end. Publisher self-regulation, a voluntary approval list, no mandatory content audit. The UK is similar, though with more detailed national curriculum specifications. The United States is the extreme case of decentralization — there is no national textbook approval process at all. Each state, and often each school district, makes its own decisions.
Which is why Texas and California effectively set the national textbook market.
Because they're the two largest adoption states. Texas alone purchases something like forty-eight million textbooks annually. If a publisher wants their book adopted in Texas, they have to meet the Texas State Board of Education's standards. And because it's not economical to produce a Texas edition and a separate edition for the other forty-nine states, the Texas version becomes the national version.
A fifteen-member elected board in Texas effectively decides what American children learn about evolution, climate change, and slavery.
We saw the consequences of that in twenty twenty-two, when a Texas social studies textbook was found to describe slavery as "involuntary relocation." That phrase made it through the publisher's internal review, through the state board's review, and into classrooms, and it was only caught because a member of the public flagged it during the public comment period.
" The linguistic equivalent of calling a house fire an unplanned thermal event.
That's the Texas model's vulnerability. It's not that the board is necessarily malicious — though some members certainly have agendas — it's that the process is political by design. The board is elected. Members run on platforms. They have constituencies. The textbook adoption process becomes a political battleground, and the losers are the students who get textbooks shaped by whichever side won the last election.
Now flip to the other end of the spectrum.
France is the centralized model taken to its logical conclusion. The Ministry of Education writes a detailed national curriculum called the programme. It specifies not just what topics must be covered, but in what order, with what emphasis, and often with what interpretation. Textbooks must conform to this programme to be approved. The ministry reviews them. The ministry can reject them.
One office in Paris decides what every French twelve-year-old learns about, say, colonialism.
That's exactly where the French model has been criticized. French textbooks on Algeria, for decades, reflected the official French state position, which was notably reluctant to describe what happened as a war of independence until nineteen ninety-nine. The term was "events in Algeria" until the French parliament officially recognized it as a war. Textbooks followed the state line.
The centralized model's failure mode is that textbooks become an arm of state narrative. When the state changes its mind, the textbooks change. When the state doesn't want to confront something, neither do the textbooks.
That's not hypothetical. Look at Japan. Japan has a textbook authorization system where the Ministry of Education reviews every textbook for factual accuracy and ideological balance. Sounds good on paper. But in twenty fifteen, the ministry required publishers to downplay the "comfort women" system — the wartime sexual slavery system — in history textbooks. Publishers were told to remove or soften references to forced recruitment. The result was a diplomatic crisis with South Korea and China, and Japanese students learning a sanitized version of their own history.
The ministry called it "factual accuracy review." Critics called it historical revisionism. And both descriptions are technically correct — the ministry was reviewing for factual accuracy as they defined it, which was shaped by a nationalist political agenda.
Which is the problem with any centralized model. The review process is only as good as the reviewers. If the government changes and appoints new reviewers, the textbooks change. In twenty twenty-three, Russia introduced a new history textbook that described the invasion of Ukraine as a "special military operation" to "denazify" the country. That wasn't a rogue publisher. That was the state curriculum, written and approved by the Ministry of Education.
Hungary's twenty twenty-two textbook portrayed Viktor Orbán as the defender of Christian Europe against Muslim immigration. China's twenty twenty-one revisions removed references to the Tiananmen Square protests. In each case, the textbook didn't just reflect the state's view — it was the state's view, printed and distributed as fact.
We have a genuine structural problem. Light-touch systems like Ireland's let ideological capture happen through market pressure. Centralized systems like France's or Russia's let it happen through state directive. Both produce biased textbooks. They just bias them differently.
Which makes Daniel's counterpoint question exactly right. If state intervention is propaganda, and market freedom is ideological capture by the loudest buyers, is there a third option?
There is, and it's expensive.
Of course it's Finland.
Finland's National Agency for Education is a quasi-governmental body with a rotating membership of educators, academics, and parent representatives. It produces a broad curriculum framework — not a detailed programme like France, more like a set of principles and learning objectives. Publishers have room to interpret, but every textbook goes through a binding review process before it can be used in schools.
The review is actually rigorous?
A twenty twenty-three study found that Finnish textbooks spend an average of eighteen months in review. Each page is reviewed by at least three subject-matter experts. The cost is approximately fifty thousand euros per textbook. For one textbook.
Fifty thousand euros to review a single book. That's not a process — that's a luxury good.
That's the catch. Finland can do this because it's a small, wealthy country with a strong institutional culture and a population of five and a half million. The entire Finnish textbook market is maybe a few hundred titles. Scaling that model to a country the size of the United States, or even to Ireland's budget constraints, is not straightforward.
The principle is transferable even if the price tag isn't. Independent review board, rotating membership, binding decisions, transparency about the review criteria. None of those things require fifty thousand euros per book.
The rotating membership is the key insight. If the same people review textbooks for twenty years, you get institutional capture. If you rotate educators, academics, and parents through the board on fixed terms, you make it harder for any single ideological faction to dominate. It's the same principle as jury duty.
The textbook jury. I actually like that framing. We trust twelve random citizens to decide whether someone goes to prison, but we trust a handful of political appointees or a publisher's marketing department to decide what millions of children learn about the world.
The Finnish model's results are genuinely impressive. International comparisons consistently rank Finnish textbooks as among the most balanced in the world. They're not perfect — no system is — but they don't have the kind of systematic bias problems we see in Ireland, Texas, Japan, or Russia.
The answer to "who watches the writers" is, in Finland, a rotating body of independent reviewers. But that brings us to the wild card in all of this. What happens when the writer isn't a person at all?
This is where it gets unsettling. In twenty twenty-five, a school district in Florida was found to be using an AI-generated history textbook. The district had purchased what was described as a "custom curriculum" from an AI platform for two thousand dollars. A parent discovered that the book contained claims that the Boston Tea Party was a protest against "high taxes on tea and sugar.
So it omitted the entire "no taxation without representation" argument? The thing the Boston Tea Party was actually about?
The AI had no concept of the constitutional argument. It had ingested a mass of text about the American Revolution and produced something that looked authoritative, used the right dates and names, and was fundamentally wrong about the meaning of the event.
The AI didn't have a bias in the political sense. It had a bias in the statistical sense — it reproduced the most common patterns in its training data without understanding what any of it meant.
This is the new frontier of the textbook authorship question. If a human writes a biased textbook, we can interrogate their motives, their sources, their editorial process. If an AI generates a biased textbook, there is no motive. There is no editorial process. There is just a statistical model that produced output that looked right.
Who's accountable? The school district that bought it? The AI company that sold it? The engineers who trained the model? The prompt engineer who wrote the query?
Currently, no one. The Florida district faced some bad press and pulled the book, but there was no regulatory framework for evaluating AI-generated curriculum materials. The district bought it because it was cheap — two thousand dollars versus the twenty to fifty thousand a traditional textbook costs — and because no one had told them not to.
The market working exactly as designed. Cheaper product, faster delivery, no questions asked.
This is not going to be a one-off. The economics of AI-generated textbooks are too compelling for cash-strapped school districts to ignore. A traditional textbook takes two to three years to develop and costs hundreds of thousands of dollars to produce. An AI can generate something that looks like a textbook in hours for a few thousand dollars. The temptation is enormous.
We're moving from a world where the problem was "who watches the writers" to a world where the problem is "what watched the writers, and can we trust it?
The AI problem intersects with the bias problem in a particularly nasty way. If you prompt an AI to write a "balanced" history textbook, what training data is it drawing on? If its training data overrepresents certain perspectives — which it almost certainly does, because that's how training data works — then its "balanced" output will reflect those overrepresented perspectives. The bias becomes baked in at the architectural level.
You can't audit an AI's editorial process because there isn't one. You can only audit its outputs, and that requires the kind of expert review that the Florida district didn't do and couldn't afford.
Which brings us back to Ireland, because the Irish Department of Education's response to the twenty twenty-five study is a real-time case study in how a country grapples with this. They announced a comprehensive review of the textbook approval process, expected to report in September of this year — so about three months from now.
The early leaks suggest what?
The leaks suggest the review will recommend moving toward a French-style centralized curriculum with mandatory content audits. The Department would specify not just what topics are covered but how they should be framed, and textbooks would need to conform before approval.
The response to finding bias in a light-touch system is to move toward the system that produced "events in Algeria" and "special military operation to denazify Ukraine.
That's exactly the debate happening in Irish education circles right now. Critics are saying: if the problem is that the current system let anti-Israel bias through, the solution is not to give the state more control over content, because what happens when the state's preferred framing on a different issue is itself problematic? What happens when a future government decides that climate change should be taught as "controversial" or that evolution should be "balanced" with creationism?
The single point of failure problem. In a decentralized system, a bad textbook can slip through, but it's one textbook. In a centralized system, a bad curriculum affects every student in the country.
The Irish review is happening in a context where similar scandals have erupted elsewhere. In the past twelve months alone, we've seen textbook controversies in Texas, Japan, Poland, and now Ireland. Poland is particularly instructive — after the change of government, the new administration has been trying to undo the nationalist textbook revisions of the previous Law and Justice party government, and it's been a massive, contentious process. You can't just flip a switch and replace every textbook in the country.
The political cycle becomes the curriculum cycle. Every election, a new set of textbooks.
Which is why the Finnish model, expensive as it is, points toward something important. It's not that Finland has no politics. It's that the textbook review process is insulated from the political cycle by design. The rotating board, the binding criteria, the multi-expert review per page — these are structural features that make it harder for any single government or interest group to capture the process.
Even that model assumes a stable institutional culture. If you transplanted the Finnish system to a country with weaker institutions or more polarized politics, would the independent board stay independent? Or would it become just another battleground?
That's the question no one has a good answer to. Institutional design matters, but institutional culture matters more. You can copy the org chart without copying the norms that make it work.
Where does that leave us? If we're answering Daniel's question directly — to what extent should textbook authorship be subject to regulation and oversight — the evidence seems to say: more than Ireland currently does, less than Russia currently does, and ideally through a body that's insulated from both market pressure and political pressure. Which is easy to describe and extremely hard to build.
The counterpoint — the argument that any state intervention is propaganda — has to contend with the fact that the absence of state intervention also produces propaganda. It just produces market-driven propaganda instead of state-driven propaganda. The Irish textbooks didn't become biased because the state intervened. They became biased because the state didn't.
The neutrality of inaction is a myth. Choosing not to regulate is still a choice, and it has consequences.
That's the thing I think most people miss about this debate. The question is never whether textbooks will reflect someone's values. They always will. The question is whose values, selected through what process, with what accountability.
If you're a parent listening to this, or an educator, or just someone who cares about what the next generation learns, what do you actually do? Because "redesign the entire institutional framework of your country's education system" is not a Tuesday afternoon project.
The most effective intervention, based on what we've seen in cases like the Irish study and the Texas textbook controversy, is to demand transparency. Ask your school board or your education department: who wrote the textbooks my children are using? What was the review process? Can I see the publisher's editorial guidelines? In Ireland, the twenty twenty-five study was only possible because a researcher spent months combing through textbooks that were publicly available but practically inaccessible — no parent was going to do that level of review. The information existed, but it wasn't transparent.
Step one is making the invisible visible. Just knowing who wrote the book and what their qualifications are would surface a lot of problems before they reach the classroom.
Step two, for educators specifically, is to teach students to read textbooks critically. If a textbook presents a one-sided view of a conflict, that's not just a problem to be fixed by getting a better textbook. It's a teaching opportunity. Why does this book say what it says? What sources is it using? What is it leaving out? What would a different textbook say? That skill — critical reading of authoritative texts — transfers far beyond any single controversy.
The best defense against a bad textbook is a student who's been taught to ask "who wrote this and why." Which is, now that I say it, exactly what Daniel's prompt is doing. He's asking the question that the system is designed to prevent you from asking.
For policymakers, the evidence pretty clearly points toward some version of the Finnish model. Independent review boards with rotating membership. Binding content standards that are specific enough to catch problems but broad enough to allow for pedagogical diversity. Mandatory expert review. Yes, it's expensive. But the alternative — leaving it to the market or to the political cycle — has repeatedly failed in ways that are also expensive, just harder to measure.
What's the cost of a generation of Irish students who've internalized that Israel is an apartheid state that commits genocide, without any context about Hamas or October seventh or the broader conflict? What's the cost of American students who learned that slavery was "involuntary relocation"? You can't put a number on that, but it's not zero.
The cost compounds. Those students become voters, journalists, politicians, diplomats. The textbook bias of twenty twenty-five becomes the foreign policy of twenty forty-five.
Which brings us to the open question that the Irish review is going to have to grapple with in September. As AI-generated textbooks become cheaper and more common, does the problem get better or worse? An AI can be programmed to be "neutral" — but who programs the AI? What training data does it use? What definitions of neutrality are baked into its prompts?
I think the honest answer is that AI makes the transparency problem more urgent and the accountability problem harder. More urgent because AI-generated content can scale in ways human-authored content can't — one bad AI textbook could reach millions of students before anyone catches the errors. Harder because there's no author to interrogate, no editorial process to audit, no motive to examine.
The black box writing the textbook for the black box of the future citizen's mind.
The Irish review, whatever it recommends, is going to set a precedent. Ireland is a small country, but it's in the EU, it's influential in education circles, and other small countries are watching. If Ireland moves to a centralized model, that strengthens the argument for centralization elsewhere. If it instead moves toward something like the Finnish independent board model, that's a different signal entirely.
Then there's Poland, trying to undo the textbooks of the previous government. Hungary, where the textbooks are currently Orbánist. The United States, where the Texas board could flip in the next election and trigger another round of textbook wars. This is not a settled question anywhere. It's an active front in a slow-moving conflict over what counts as knowledge.
The phrase "written in the history books" is not a metaphor. It's a process. It's happening right now, in publisher offices and ministry conference rooms and school board hearings, and most of us never think about it until a study like the Irish one surfaces something shocking.
By then, the books have been in classrooms for years.
If you want to understand how ideas become facts — how a particular view of the world gets elevated to the status of "what everyone knows" — this is the machinery. These are the people. And the question of who watches them is not academic. It's about what your children and grandchildren will believe is true.
Now: Hilbert's daily fun fact.
Hilbert: In the 1970s, Nepalese mathematics textbooks used a unique notation system where the number one lakh — one hundred thousand — was written with a single character resembling a stylized spiral, a holdover from an earlier mercantile shorthand that treated large units as indivisible wholes rather than decimal aggregates. This meant that a Nepalese student solving a problem involving one hundred thousand items would write one character where a Western student would write six digits, effectively performing a unit conversion with every calculation.
A spiral for a hundred thousand. That's actually... kind of elegant.
I'm now imagining doing my taxes in spirals.
This has been My Weird Prompts, produced by Hilbert Flumingtop. If you want to understand how ideas become facts, this is the episode that shows you the machinery — and the machinery is people, sitting in offices, making decisions most of us never think to question. You can find every episode at myweirdprompts dot com or wherever you get your podcasts. If you found this worth your time, leave a review — it's how more people find the show. We'll be back with another one soon.
