When to Use AI for Learning and When Not To with Linda Berberich, PhD
From Transformative Principal · Hosted by Jethro Jones
Episode summary
In this episode of Transformative Principal: Linda Berberich, a behavioral scientist who runs Linda B Learning in Seattle, has built machine learning for training since before AI became a marketing term. She explains how to tell which learning tasks suit AI in education, using Ruth Clark's near transfer and far transfer tasks, and why a computer is better at some than others. Her rule: use computers as practice partners for skills with a clearly right answer, like oral reading fluency and math facts, because…
Original publisher description
Linda Berberich, a behavioral scientist who runs Linda B Learning in Seattle, has built machine learning for training since before AI became a marketing term. She explains how to tell which learning tasks suit AI in education, using Ruth Clark's near transfer and far transfer tasks, and why a computer is better at some than others. Her rule: use computers as practice partners for skills with a clearly right answer, like oral reading fluency and math facts, because feedback is always accurate. GPTs can give different feedback to the same input and flatter weak work. Jethro shares a doctoral paper he wrote twice with two different GPTs. What you'll learn Use computers as practice partners for skills with a definite right answer, such as oral reading fluency and math fact memorization.Don't expect a GPT to give consistent feedback; Berberich says the same interaction can return different responses.Check AI praise skeptically, since GPTs can tell you a weak response was great.Match any technology to the culture and the people in your school as they are now, then keep adjusting.Let students lead with technology, and start from interests such as dance to draw them into STEM. Chapters 1:20 AI as machine learning, not a buzzword4:24 What technology is really good at doing9:52 Near transfer, far transfer and organic skills13:56 When to use computers as practice partners15:09 Why GPT feedback is inconsistent and flattering19:58 Jethro writes one paper with two GPTs25:24 Fit technology to your school's culture27:39 Generative instruction and personalized learning Questions this episode answers What learning tasks are best for a computer or AI? Berberich says computers excel at tasks with a definite right and wrong answer, such as oral reading fluency and math fact memorization. The computer gives accurate, consistent feedback, while a human practice partner may look away or nod along. Complex, creative tasks need a different approach. What is the difference between near transfer and far transfer? Berberich, citing Ruth Clark, says near transfer tasks are procedural and done the same way every time, like making a Starbucks latte. Far transfer tasks are strategic: what you do varies with context, like a sales call where the customer's response changes your next move. Why can't you trust GPT feedback for learning? Berberich says GPTs are not consistent: run the same interaction twice and you get different feedback. Recent models also pander, praising weak responses. Jethro adds that his GPT trained on his voice made more mistakes and hallucinated more than the untrained one. About Linda Berberich, PhD Behavioral scientist and runs the Linda B Learning consultancy in Seattle, specializing in learning science and intelligent technology product design. Links and resources Ruth Clark on far transferPaper written with two GPTsLinda B Learning Working on this at your school? Get a free preview of a 10-week leadership pathway built from episodes like this one at transformativeprincipal.com/start. Thanks to our Sponsor Broadcast Made Easy who help you start a news broadcast at your school by having everything you need in one easy-to-download place. Check it out today and get your kids creating. We’re proud to be sponsored by IXL. IXL’s comprehensive teaching and learning platform for math, language arts, science, and social studies is accelerating achievement in 95 of the top 100 U.S. school districts. Loved by teachers and backed by independent research from Johns Hopkins University, IXL can help you do the following and more: Simplify and streamline technologySave teachers’ timeReliably meet Tier 1 standardsImprove student performance on state assessments 🚀 Ready to see why leading districts trust IXL for their educational needs? Visit IXL.com/BE today to learn more about how IXL can elevate your school or district.