Murphy: AI Use In Classrooms Needs a Lot More Scrutiny

WASHINGTON - U.S. Senator Chris Murphy (D-Conn.) this week, during a hearing of the U.S. Senate Health, Education, Labor, and Pensions (HELP) Committee, discussed the risks of rushing AI into schools without a plan to regulate the technology or verify whether AI improves the educational process. Murphy expressed serious doubts that AI companies have the right incentives to support student learning.

Speaking as a parent, Murphy noted AI's corrosive effects on kids' mental and social health: "I don't think there's any way to overhype the damage that AI has already done to our kids' ability to succeed. I'm the parent of an incoming 12th grader and incoming 9th grader, and I've seen already how, at scale, kids have outsourced critical thinking, have outsourced friendship, have even outsourced moral advice to AI."

Murphy laid out why we shouldn't blindly trust that AI companies are working in society's best interests: "The AI companies have a veneer of social benefit, but an underlying motivation of addiction and profit expansion... One observer of the major tech companies said these companies would burn everything to the ground as long as they're making a profit."

Murphy called for common sense guardrails around introducing AI into the classroom: "We require a teacher to go through licensing before they are in a classroom. We require most complicated technologies to go through an approval process before they are handed out to the American public. What do you think about an idea to make sure that AI technology that is being handed to a teacher go through some expert approval process first to make sure that its actual function is to supplement the teacher's instruction, rather than its primary function is to try to spin you into higher usage and higher pay rates for the company that owns it?"

The full transcript of Murphy's exchanges is available below.

Sen. Chris Murphy: I don't think there's any way to overhype the damage that AI has already done to our kids' ability to succeed. I'm the parent of an incoming 12th grader and incoming ninth grader, and I've seen already how, at scale, kids have outsourced critical thinking, have outsourced friendship, have even outsourced moral advice to AI completely unregulated interactions between machines and young children, something we would have never thought of allowing just decades ago. All your testimony was compelling, but Ms. Mote, I thought yours was especially compelling, given your talk about this idea of cognitive surrender that can happen. Let me maybe ask this question about cognitive surrender first. So, I like that you're talking about AI as a tool for teachers, not just unleashing it on kids as guinea pigs, but the AI companies have a veneer of social benefit, but an underlying motivation of addiction and profit expansion. This is from a chilling New York Times expose this weekend regarding the unleashing of deep fakes without any controls on the American public. One observer of the major tech companies said these companies would burn everything to the ground as long as they're making a profit. So, let's say we're successful in just having this pointed at teachers as a tool, but if the AI companies are only interested in selling more, in having teachers use it more. How do you protect against teachers being sucked into some kind of algorithmic addictive product, like kids are, if the intent is not social benefit, but the product being used for more and more and more. How do you make sure that teachers aren't being sucked in like kids are being sucked in?

Ms. Erin Mote: Thank you for the question. I will want to give full credit to professors at Wharton who came up with the study around cognitive surrender and this idea that we are offloading tasks and offloading the discernment that Senator Kaine just talked about when he looks at those AI summaries. So, there's two things I think that are important, Senator Murphy, when we think about students versus teachers. The first is what neuroscience tells us. There's a really big difference in putting this technology in front of a brain that is developing and is doing that offloading, frankly, in a way, where those neuro pathways are still developing. The difference between Senator Kaine's experience and that of a young person is that Senator Kaine has the ability to read what his friends are saying, knows their preferences, and has those neuro pathways really concrete. So, I think there is a different set of risk with young people than there is with folks who, frankly, are over the age of 25. If I had the rules, I would say it's rental car rules, 25 and over, because that's really when we know the brain has gelled in terms of those neural pathways. But we also need to invest in professional development for educators. Here's what we know about actual effective uses of AI: is that when AI amplifies human potential, when the human alone is working with a student, learning can go up about 4.5 percentage points, but when the AI is providing that just-in-time data diagnostic, when it's potentially giving a hint with that judgment, students grow 10 percentage points with the same tool. So, it really is, how is AI supporting that human-centered knowledge transfer? That's the type of work we need to be focusing on in education.

Murphy: But, I think it might be a little naive to suggest that an adult is magically immune to these incredibly strong brain worm-like algorithms. I agree with you that they are more immune than children. We require a teacher to go through licensing before they are in a classroom. We require most complicated technologies to go through an approval process before they are handed out to the American public. What do you think about an idea to make sure that AI technology that is being handed to a teacher go through some expert approval process first to make sure that its actual function is to supplement the teacher's instruction, rather than its primary function is to try to spin you into higher usage and higher pay rates for the company that owns it?

Mote: Thank you for that question. We're developing a learning sciences benchmark, actually, right now with Arizona State University to measure consumer-grade AI tools and the impact that they have on learning and cognition, to give educators and decision makers an idea of how a tool stacks up. For me, the work that Delaware is doing, Utah is doing, even Vermont is doing in creating sandboxes at the state level to test these tools to build evidence and to make that evidence digestible to educators is a huge improvement on how we're going to be able to make sure that all schools have access to information about what works for whom and under what conditions.

Murphy: You were nodding your head, Secretary Marten. Just wanted to give you a chance to respond.

Secretary Cynthia Marten: I'm very much nodding my head at this, because you're asking to interrupt an algorithmic cycle. The one that can interrupt - that is the human. We can actually see it for what it is, realize we're being taken down a rabbit hole, and stop it, but that's at the individual level. You're asking at a systemic level, nationally, statewide, districtwide. We want that cycle interrupted, and it has to do with what is being demanded as the outcome of public education. What do we want our students leaving knowing and being able to do so a teacher is not going down a rabbit hole that's not producing the outcome of student learning? We want critical thinking. We want kids to be able to design and defend their point of view and their perspective on something. No tool is going to take you - you're in some rabbit hole trying to figure out the actual, not the problem that you're trying to solve, which is student learning. I want to know, did students learn? Was I teaching, and did students learn?

Murphy: And I appreciate it. I'm over time, Mr. Chairman, but I imagine that a product, right, that is interested first in return to investors, are for instance going to point you to the purchase of other products that they make, right, that might have nothing to do with student outcomes, but are certainly aligned with the profit return, and so I just want to make sure we really think through what the underlying motivations are of the people who are making the software before we just assume that their motivations are the benefit of our kids. Thank you, Mr. Chairman.