TeachLab with Justin Reich

Teachers have all sorts of opinions about AI. Some are optimistic, some are pessimistic. But the most common topic that came up in our interviews was cheating.

While students have always taken shortcuts to complete their work, ChatGPT and other generative AI have a historically unique power to quickly, convincingly and comprehensively do a students’ assignment. This is proving a powerful temptation to students.

So how do teachers help their students make good decisions? Teachers know that schools have historically struggled to manage discipline fairly but they also recognize that letting students get away with cheating isn’t doing them a favor. Teachers share how they’re navigating the Scylla and Charybdis of school discipline in the AI age.

Listen to a bleeped version of this episode (Coming soon!).

Transcript coming soon!

This episode was produced by Jesse Dukes with Yebu Ji.
Editing: Alexandra Salomon and Ruxandra Guidi
Reporting and research from Natasha Esteves, Andrew Meriwether, Holly McDede, Andrew Parsons, Marnette Federis, and Chris Bagg.
Sound design and music supervision by Steven Jackson.
Production assistance from Nathan Ray.
Data analysis from Manee Ngozi Nnamani and Manasa Kudumu.
Special thanks to Josh Sheldon, Camila Lee, Liz Hutner, and Eric Klopfer.
Administrative support from Jessica Rondon.

Thanks to the teachers who spoke to us including Joe O'Hara, Alec Jensen, Schuyler Hunt, Anna Rose Pandey, Ray Salazar, and Jessica Petit-Frere. And thanks to all the teachers and students who partipated in our research.

Thanks to Greer Murphy and Jessa Kirk, at UC Santa Cruz's Office of Academic Integrity. Check out Greer Murphy's co-authored survey of academic integrity policies.

The research and reporting you heard in this episode was supported by the Spencer Foundation, the Kapor Foundation, the Jameel World Education Lab, the Social and Ethical Responsibility of Computing initiative at MIT, and the RAISE initiative, Responsible AI for Social Empowerment and Education also at MIT.

We had support from Google’s Academic Research Awards program.

The Homework Machine is a program of the MIT Teaching Systems Lab, Justin Reich, director.

ChatGPT is the most well known of the Large Language Models (LLMs) but what is an LLM? We go deep into how this remarkable new technology is built, and why their performance is inconsistent — or jagged — across similar tasks. We dive into the techniques AI engineers use to align these tools’ behavior with our values, and explain why they don’t always work, and sometimes we get hallucinations or biased output. 

This episode was produced by Steven Jackson and Jesse Dukes

Editing:  Alexandra Salomon and Ruxandra Guidi  

Reporting and research from Holly McDede, Natasha Esteves, Andrew Parsons, Andrew Meriwether, Marnette Federis, and Chris Bagg.

Sound design and music supervision by Steven Jackson. 

Production assistance from Yebu Ji and Nathan Ray. 

Data analysis from Manee Ngozi Nnamani and Manasa Kudumu. 

Special thanks to Josh Sheldon, Camila Lee, Liz Hutner, and Eric Klopfer. 

Administrative support from Jessica Rondon. 

The research and reporting you heard in this episode was supported by the Spencer Foundation, the Kapor Foundation, the Jameel World Education Lab, the Social and Ethical Responsibility of Computing initiative at MIT, and the RAISE initiative, Responsible AI for Social Empowerment and Education also at MIT. 

We had support from Google’s  Academic Research Awards program. 

The Homework Machine is a program of the MIT Teaching Systems Lab, Justin Reich, director. 

In late November of 2022, ChatGPT was released to the public as a free research preview. Students quickly realized ChatGPT was pretty good at doing their homework for them. Schools scrambled to figure out what to do:  Ban it? Embrace it? Teachers and students found themselves adapting to a new reality. 

Buckle Up, Here it Comes kicks off “The Homework Machine” a mini series in the Teachlab podcast. Hosts Jesse Dukes and Justin Reich share stories of teachers and students reacting to the arrival of an exciting, alarming, and strange new technology. 

Producer: Jesse Dukes

Editors: Ruxandra Guidi and Alexandra Salomon.

Reporting and research: Holly McDede, Natasha Esteves, Andrew Meriwether, and Chris Bagg. 

Sound design and music supervision: Steven Jackson. 

Data analysis: Manee Ngozi Nnamani and Manasa Kudumu. 

Special thanks to Josh Sheldon, Camila Lee, Liz Hutner, and Eric Klopfer. 

Administrative support from Jessica Rondon. 

The research and reporting you heard in this episode was supported by the Spencer Foundation, the Kapor Foundation, the Jameel World Education Lab, the Social and Ethical Responsibility of Computing initiative at MIT, and the RAISE initiative, Responsible AI for Social Empowerment and Education also at MIT. 

Additional support from Google’s Academic Research Awards program. 
InTandem facilitated some of our student interviews. 

Full episode transcript. 

Generative AI is not like other education technologies, which schools often invite into the classroom. This one crashed the party. And then, it started re-arranging the furniture. We wanted to learn more, so in a little over a year, the Teaching Systems Lab has talked to over 90 teachers and 30 students about the impact of Generative AI. 

Some are excited about AI's potential to transform education for the better. 

Others are troubled by the temptations of a machine that can quickly and convincingly do many homework assignments. 

And some think AI is just a shiny object – a distraction from the much larger problems facing education.

Over the next seven episodes, we'll try to answer the question: Is AI a game-changing tool, a threat to critical thinking, another ed tech fad? … or something else? 

Media requests can go to jpd009@mit.edu. 
 

School leaders, education researchers, and others often point to a study conducted by Stanford researchers that suggested the arrival of generative AI in K-12 school has not meaningfully increased the percentage of students who acknowledge some kind of academic dishonesty. Sometimes, school leaders or experts suggest it means there’s “nothing to see here” when it comes to tools like ChatGPT and worries about students bypassing learning. Researcher + Journalist Jesse Dukes joins Justin to dive into the specifics of that study, and compare it with anecdotes from interviews with students and teachers. 

 This episode was produced by Andrew Meriwether and Jesse Dukes. We had additional reporting from Holly McDede  and research help from Natasha Esteves and Manassa Kudumu. Thanks to the Spencer Foundation and the Social and Ethical Responsibilities of Computing Initiative at MIT for funding our ongoing  research into the arrival of generative AI in schools. And thanks to the Kapor Foundation for funding Jesse’s work in California with KALW public radio. Thanks to all of the teachers and administrators who have talked with us. 

 If you want to take our survey, or learn more about our research into Generative AI and K12 education, head over to tsl.mit.edu/ai and if you want to volunteer for the sister study, in California, visit Jesse’s Linktree.

Justin Reich and researcher and producer Jesse Dukes argue that AI in requires a new theoretical framework. Generative AI, unlike many teaching technologies, is an "arrival" technology, meaning it will be present in school environments regardless of what choices school leaders make about whether to adopt it. Their new preprint Toward a New Theory of Arrival Technologies: The Case of ChatGPT and the Future of Education Technology after Adoption explores the idea of "arrival technologies"

But rather than summarize it, hey, why not let the arrival technology do it? So we turn to Google's new tool, Notebook LM for a "Deep Dive Conversation" to summarize the article. 

Thanks to the Spencer Foundation and the Social and Ethical Responsibilities of Computing initiative at MIT for funding our ongoing  research into the arrival of generative AI in schools. Thanks to all of the teachers and administrators who have talked with us.  This episode was produced by Jesse Dukes. We had research help from Chris Bagg, Manasa Kudumu, Natasha Esteves, and Andrew Meriwether. If you want to take our survey, or learn more about our research into Generative AI and K12 education, head over to tsl.mit.edu and if you want to volunteer for the sister study, in California, visit Jesse’s Linktree.

Jesse Dukes reports from a two day training one school district offered its teachers, all to help them adapt to the arrival of generative AI in schools. That training proved helpful to teachers, but it’s not clear how much professional development most American teachers are getting, and it appears many are getting little or none. 

Episode produced by Jesse Dukes

Research by Chris Bagg, Andrew Meriwether, and Natashas Esteves. 

Editorial support by Natasha Esteves. 

Additional research from Manasa Kudumu. 

Thanks for the school district that let us visit (you know who are!) 

Thanks to all the teachers who talked with us for our ongoing research into the arrival of generative AI in school environments. Learn more at tsl.mit.edu/AI. 

We've just wrapped up a school year, so our team researching the arrival of generative AI in classrooms shares  some fun and inspiring moments of teachers adapting to the new reality. We hear from teachers who role modeled the use of as a tool or resource for students, or to generate amusing and weird curricula. 

This episode was produced by Andrew Meriwether and Jesse Dukes, and features research by Natasha Esteves, Chris Bagg, Andrew Meriwether, and Jesse Dukes. Original song "The AI is a Hummin'" written by Jesse Dukes with help from Bing CoPilot, and performed by Jesse Dukes. 

To learn more about our ongoing research, take a survey, or volunteer to be interviewed, visit https://tsl.mit.edu/AI/.

The Arrival of AI powered tools like ChatGPT (now GPT4) in schools has generated concerns that students would use the tool to bypass cognition, or, “cheat” as we colloquially call it. And, it appears many students are doing just that. Early on in our research project about generative AI’s arrival in schools, it seems that English, ELL, and Social Studies at upper levels are particularly likely to encounter students using AI, without permission, to complete assignments. Justin talks to Jesse Dukes who shares a few examples of how teachers are managing this new challenge, and they discuss the wider implications for teaching, ethics, and student well being. 

This episode was produced by Jesse Dukes. We had research help from Natasha Esteves, Andrew Meriwether, and Chris Bagg. Thanks to the Spencer Foundation for funding our research, and to all the teachers and administrators who agreed to speak with us.