Data Science Leaders Transforming Education with Generative AI and Active Learning
15 snips
Nov 8, 2023 
Stephen Kosslyn, cognitive scientist and education innovator who led programs at Harvard, Stanford, and Minerva, and now heads Active Learning Sciences. He discusses AI as a cognitive amplifier. He outlines active learning by doing, scalable AI-driven role-play and assessments, personalization and deliberate practice, plus tools to prevent misuse and protect student data.
AI Snips
Chapters
Books
Transcript
Episode notes
Variation In AI Responses Can Be A Feature Or A Bug
- AI outputs vary per run, which can be pedagogically useful or problematic depending on control needs.
- Kosslyn recommends either iterating as a designer to prepackage exact interactions or embracing variation for student exposure.
Don't Use LLMs As Search Engines Use Them For Practice
- Avoid using generative AI as an information retrieval system; instead, use it to design and run active practice and debriefs.
- Structure learning as a "learning sandwich": front-loaded content, AI-driven use, then AI-guided debrief with rubrics.
Use AI Randomness To Enable Varied Deliberate Practice
- Exploit AI randomness to give students varied practice that approaches the same learning objective from multiple angles.
- Use debriefs to identify rough spots and focus deliberate practice on those weaknesses.
