Episode 81

AI, History, and the Fight for Source Truth

Guest name: Jan van der Crabben
Guest title: Founder & CEO
Guest company: World History Encyclopedia
Host: Eric Stano

Brief description of the episode

AI makes it dangerously easy to get “one confident answer” and stop there. For education publishers and learning platforms, that’s a product problem: teachers want sources they can trust, students need support for research (not shortcut cheating), and your credibility can’t survive a black-box answer engine.

In this episode, Jan van der Crabben (World History Encyclopedia) shares how the encyclopedia was built around a simple idea: history is a web of connections, not isolated timelines. He explains why history education is uniquely powerful for teaching source evaluation, and why AI in education needs guardrails, citations, and context-aware design.

Jan also breaks down how their History AI tool evolved based on teacher feedback, why shorter answers and prominent citations matter, how adding academic journals can reduce hallucinations when your own corpus is thin, and why the future of trustworthy publishing is at stake if AI becomes the default interface to knowledge.

Key Takeaways:

- Students learn more when they can follow connections across places and time, not just memorize dates.
- Digitizing history works best when your content system is built to surface relationships automatically.
- The learning experience improves when curiosity is rewarded with “next questions” and linked sources.
- History trains source evaluation: who created the information, what they wanted, and how reliable it is.
- That same skill is how students should approach AI-generated answers and social media claims.
- If you want informed participation in society, you need better “trust judgment,” not just more information.
- A handful of systems can end up controlling how the world retrieves information.
- AI makes it easier to accept a single synthesized answer without encountering other perspectives.
- If fewer people visit trusted sources, publishing becomes less viable, and the pipeline for reliable knowledge weakens.
- Teachers asked for shorter answers with stronger visibility into sources, not long chatbot-style responses.
- The product was shaped by educator testing and refined around classroom realities.
- The tool includes guardrails to avoid inappropriate prompts, non-history prompts, and essay-writing behaviors.
- Separate student use from teacher prep (teachers may need longer lesson-planning outputs).
- Treat AI like a research assistant: useful, but always something to check, not a final authority.
- Minimize data, separate identity from analytics where possible, and design with school privacy expectations in mind.
- World History Encyclopedia: https://www.worldhistory.org/
- CORE (Open University open research index, referenced in the conversation): https://core.ac.uk/
- Further reading: AI in schools: what to do before picking a tool

AI, History, and the Fight for Source Truth

Brief description of the episode

Key Takeaways:

You may also like this

Get the Citation‑First AI Checklist