AI is everywhere. NYC schools want to keep it away from young kids

**SPEAKER_1** (0:00)
Support for Here & Now Anytime comes from MathWorks, creator of MATLAB and Simulink software, to design and develop engineered systems, accelerating the pace of discovery in engineering and science. Learn more at mathworks.com.

**Joseph Winters** (0:15)
WBUR Podcast, Boston.

**Zeran Mamdani** (0:21)
Kids need to learn to think for themselves. They need to work alongside their peers. They need to learn how to ask better questions, and they need to learn what it feels like to not know the answer.

**Chris Bentley** (0:32)
Getting AI out of the ABCs. It's Wednesday, September 2nd, and this is Here & Now Anytime from NPR and WBUR. I'm Chris Bentley.
Today on the show, the fossil fuel and petrochemical industries know they've got a bad rap for plastic pollution. Part of their solution is pushing curricula for school kids about the positive sides of plastics, PR, or propaganda. Also, the mobile game Candy Crush makes a billion dollars in a typical year. How did it get so profitable?

**Alanna Okun** (1:15)
Industry analysis suggests that only 4 percent of players ever spend money, and that's millions and millions of players. So, you know, you can do the math there.

**Chris Bentley** (1:24)
Before we get to that story though, New York kids and teachers are going old school. For one year, the city will ban the use of artificial intelligence for its youngest students in public schools. Kids won't be allowed to use the technology until they enter high school, and even then there will be some restrictions. The new rules also limit screen time and stop schools from handing out phones and tablets to students until they reach third grade.
Mayor Zeran Mamdani said it will apply to about 600,000 students and their teachers.

**Zeran Mamdani** (1:56)
Kids need to learn to think for themselves. They need to work alongside their peers. They need to learn how to ask better questions, and they need to learn what it feels like to not know the answer.
This is what learning is all about, and it is what city government is determined to protect.

**Chris Bentley** (2:13)
Jill Barshay is with The Hechinger Report, a nonprofit education news outlet. She told Scott Tong more about how this will work.

**Jill Barshay** (2:21)
There are a lot of exceptions, and then there are a lot of questions about implementation. The fine print says that centrally approved instructional software can still have AI, including e-books and coding software.
Standardized tests can still use AI. Students with disabilities can still have AI. English learners can. And teachers are allowed to use AI for lesson plans, translating from foreign languages, and even writing messages to parents and colleagues.

**Scott Tong** (2:57)
Wow. Okay. So it is complicated, as you're saying. Are there any bright lines, say, whether for teachers or students, certain uses or certain kids of a certain age where the answer is no?

**Jill Barshay** (3:12)
Well, the one thing that the regulations are saying is teachers cannot use AI for grading, and they can't use it for deciding who will graduate, and they're not allowed to use it to help children with mental health issues. So there are some specific bright lines in it. The other thing that's really interesting is that so many educational apps have already embedded AI in them. Google has embedded AI in its product. Power School, which is used by many schools in New York City, has embedded AI there.
There are many apps that use it. Khan Academy has AI embedded in it. And so it's interesting how that's going to work in practice. Will the companies take AI out, or will it sort of slip through anyway?

**Scott Tong** (4:09)
Yeah. So I guess you're suggesting, how is this going to be implemented? How is this going to be enforced? Those are still big questions out there for New York City.

**Jill Barshay** (4:19)
Exactly.

**Scott Tong** (4:21)
Yeah. Yeah. Now, I mean, schools everywhere in this country beyond New York City are debating AI. Jill, you attended a number of education conferences in California.
What did you glean about the nature of this conversation, the contours of the debate?

**Jill Barshay** (4:39)
I think we need to distinguish how AI can be independently used by students using commercially available AI apps. So for example, a student can go to ChatGPT or Claude or Gemini and have it do its work for them.
We've seen well-designed rigorous studies showing that this is harmful to learning. When AI does the work for you, you don't learn and there have been really well-designed randomized control trials showing, say you use AI to help you with homework and then later you take a test. Those students do much worse on the test. They're not building the skills. But what's complicated about it is we're also seeing researchers come up with very specific uses of AI with a lot of guardrails, and they're experimenting and using AI in different ways, and they're starting to see some benefits on how it might be able to improve learning. There was one experiment recently in Tennessee where it was used to help students go over their mistakes in math and slow things down, and look at the problem step-by-step, and there was actually a learning benefit to it.