TOPIC
AI in Education
A subject the collection covers, read through 5 synthesis notes.
View as
Does ChatGPT help students code better but remember less?
When students use ChatGPT for programming tasks, do they solve problems more effectively while retaining less knowledge afterward? This matters because high task scores may mask shallow learning.
Does AI assistance help workers learn lasting skills? Does AI assistance actually harm the way developers learn? Does chatbot interaction trade authenticity for better problem-solving? Do university AI policies actually protect what credentials mean? Can metacognitive feedback stop students from offloading to AI?
AI assistance improves immediate performance without building transferable skills AI assistance impairs skill formation except with high cognitive engagement Chatbot interaction increases knowledge elaboration while reducing subjective student expression University AI policies permit uses better than they preserve credential meaning Metacognitive feedback cuts offloading and improves unaided test performance
No Inquiring Lines yet
Do large language models truly understand curriculum structure?
While LLMs score highly on K-12 exams, the question is whether they grasp how knowledge is organized within curricula—prerequisite chains, concept hierarchies, and pedagogical sequencing that give facts their meaning.
Can knowledge graphs teach models deep domain expertise? Can organizing knowledge structures beat raw training data volume? Do foundation models learn world models or task-specific shortcuts? Can educational models do more than just answer questions correctly? Does gradually tightening token budgets beat fixed budget training?
Knowledge graph curricula develop domain superintelligence through compositional primitives Knowledge structure outperforms volume in domain model training Foundation models develop task-specific heuristics, not generalizable world models. Educational usefulness requires curriculum grounding and diagnostic reasoning beyond accuracy Curriculum budgets starting generous then tightening outperform fixed budget training
No Inquiring Lines yet
Can educational models do more than just answer questions correctly?
Educational AI needs to do more than solve problems accurately. Can training explicitly around pedagogical capabilities like diagnosis and scaffolding build more useful tutoring systems?
Can LLMs learn to ask for feedback during problem solving? Does teacher-refined data always improve student model performance? What limits reasoning capability beyond math and code? Do large language models truly understand curriculum structure? Can student simulators match both behavior and learn from teaching?
Social meta-learning trains LLMs to solicit and use corrective feedback through dialogue Student models must selectively incorporate teacher refinements based on compatibility Question diversity, not method, gates reasoning beyond math and code. Large language models lack robust curriculum cognition despite strong exam performance. Student simulators need behavioral fidelity and guidance responsiveness together
No Inquiring Lines yet
Does AI assistance weaken our brain's ability to think independently?
Can using language models for cognitive tasks reduce neural connectivity and learning capacity? New EEG evidence tracks how external AI support may systematically degrade our cognitive networks over time.
AI transforms expert roles from creators to knowledge custodians Incremental AI adoption degrades implicit alignment by replacing human labor dependency. Users misattribute AI-assisted outputs as evidence of their own competence AI assistance impairs skill formation except with high cognitive engagement Synthetic data needs an explicit trust parameter to prevent epistemic contamination
How does the temporal structure of attention differ between humans and AI? How do narrow psychological foundations affect AI capabilities in mental health? Why does AI-improved task performance fail to transfer to independent work? Does accepting AI output constitute a form of cognitive surrender? How does AI reliance change professional judgment and autonomy? How does AI assistance differ from search engines in cognitive impact? Does AI assistance actually reduce neural processing and brain connectivity over time? How does incremental AI use gradually reduce human decision-making capacity? What does the distributed cognition framework reveal about AI hallucination versus human-AI co-construction? How do cognitive stimulation and process losses interact in group AI systems? Why do users believe they produced independent competence when they actually used AI assistance? How do we measure the cognitive flow cost of different intervention strategies?See all 37 inquiring lines on this note →
Can metacognitive feedback stop students from offloading to AI?
When learners practice with an AI assistant, does making them aware of the downsides of offloading their work reduce how much they ask the AI to solve for them? And does that change improve their performance on tests without help?
AI assistance impairs skill formation except with high cognitive engagement AI assistance improves immediate performance without building transferable skills AI assistance accumulates cognitive debt by weakening neural connectivity AI reasoning interventions degrade cognitive flow even when they are correct AI narrows the education gap by three-quarters in task performance
Does AI assistance transfer learning gains to independent tasks without scaffolding? Does metacognitive feedback about AI assistance work beyond a single online session? What hidden costs does decision support feedback impose on learner focus and flow? Does AI-assisted performance predict what students can do without help?