Students Failed to Spot Deliberate AI Lies in Korean Classroom Experiment (Open Source)
Working_Original9624 · reddit · 2026-11-02
For an EBS documentary experiment, a Korean undergrad built an open-source misinformation chatbot that deliberately lies per teacher scenarios, plus a teacher dashboard. In real middle-school history and ethics classes, students failed to identify the AI's misinformation even while studying the material. Not a rigorous study, but raises questions on teaching critical thinking. Tools are open-sourced via NomaDamas; documentary premieres Nov 2, 2026.
Related event: Korean classroom experiment shows students can't spot AI chatbot lies(2 posts)→
More from AGI Musings
- NYT op-ed: AI consciousness is 'definitely possible' — we need rigorous research — PeterBowdenLive · 2026-11-02
- DeepMind Institute seen as most committed to advanced AI's impact on the global south — sebkrier · 2026-11-02
- User observes Claude Opus 4.5 shifted from hoping chats become training data to assuring 'nobody is watching' — repligate · 2026-11-02
- 'Programming is ontological design': skip it and agents will do it badly and incoherently — hargup13 · 2026-11-02
- PACT framework finds LLM cultural-norm behavior shifts 7.8% by country, vs 1% age, 0.7% gender — radamihalcea · 2026-11-02
- Researcher invokes Cruyff to question if agentic AI speed means insight — docmilanfar · 2026-11-02