Education: What A.I. Looks Like in Practice

Before I came to see the Seckinger schools, I struggled to understand what embedding A.I. in every grade would actually look like. In my dark fantasies, I pictured kindergartners having conversations with chatbots instead of one another. I imagined older children plugged into laptops doing an A.I.-personalized set of math problems, while a human teacher sat off in the background, marginalized by bots.

It’s definitely not that. Educators were not sidelined. I was struck by the warmth of all the teachers I observed. I saw a first-grade teacher, Liza Earley, running a literacy activity that was in line with the green “mathematical reasoning” triangle in the A.I. framework. Ms. Earley was showing her students how to observe patterns in words that they hear, like “high,” “tie” and “fly.” She noticed that a few kids were not picking up on the relationship among the words, so she crouched down to their eye level on the rug and gave them further guidance.

In middle school, the activities started using technology I recognized as artificial intelligence. An eighth-grade social studies teacher at Jones Middle School, Kacie Holycross, joked with her kids while they posed as journalists interviewing chatbots Ms. Holycross built on MagicSchool, an A.I. platform for K-12 education.

The chatbots were supplied with biographical information about Richard Russell Jr. and Carl Vinson, two long-dead Georgia politicians who brought lots of military funding to the state during World War II. Ms. Holycross wrote the questions for the children to feed the chatbot. Then the Vinson bot spat out replies like “Georgia’s got the workers, the shipyards and the know-how to build them.”

Ms. Holycross told me that interacting with primary sources using A.I. made the lesson more “student friendly.” But the kids I observed did not seem especially riveted. I asked one student if he planned to ask the bots any follow-up questions. “I could do that,” he said, but didn’t.

Ms. Holycross told me, “Their goal right now is just to kind of focus on the three questions and then summarize what they say.” Some of the students need structured guidance with the chatbots, she explained to me later.

My last stop was Seckinger High School, which is the fanciest public school building I have ever set foot in. I toured state-of-the-art mechanical engineering classrooms and glass-walled student “collaboration rooms.” I spoke to four teachers, three students and Seckinger’s principal about their experiences teaching and using A.I. at Seckinger.

Scott Gaffney, who is head of the social studies department, talked about how students in his A.P. human geography class used a large language model to gather data about a 2014 “snowmageddon” that paralyzed Atlanta. Mr. Gaffney and one of his students described looking up road density data and the number of students who needed to be bused home from school. The kids were instructed to come up with public policy solutions that might have made the storm less debilitating.

Mr. Gaffney said that this year the activity was more efficient than when he first taught it. Now, “kids can look at this data and they’re like, oh, 500,000 without power, 25 people dead, two inches of ice, Atlanta paralyzed, like kids stuck at school,” he explained. The students move more “obviously to a faster conclusion.”

These are remarks similar to the ones Mr. Gaffney made in an interview with CBS News in 2023. Before A.I., he could have had students look up years of traffic data, but it would have taken four to five days for the students to complete the lesson. Few people I talked to when I visited this year seemed to think that there had been any value in that slower kind of research or that there might be cognitive or emotional muscles involved in more painstaking, frustrating work that would atrophy from disuse.

As the day wore on, I kept hearing about how the students used algebraic reasoning to think more deeply about literature, or rebranded simple patterns in words as “algorithms.” They kept circling back to those colorful triangles — ethics, creative problem solving — that didn’t necessarily have anything to do with artificial intelligence.

A visual art teacher, Megan Fowler, described encouraging her students to use ChatGPT to help them free-associate words to get past artists’ block. But wouldn’t they be just as well served by thumbing through a book of paintings, going outside and looking around, or daydreaming?

I thought about Mr. Gaffney’s description of his students moving rapidly through their research assignments, and about introducing the term “user experience” to first graders. Gwinnett County often seemed to be transposing the language and goals of tech corporations onto its schools. In one sense, these goals are not new. The promise of tidy, industrialized efficiency has been central to the argument for bringing tech into schools for at least a hundred years.

In 1913, Thomas Edison predicted that books would quickly become obsolete: “Scholars will soon be instructed through the eye. It is possible to touch every branch of human knowledge with the motion picture.” This is just one example of utopian educational technology boosterism that Larry Cuban, who is a professor emeritus at Stanford, includes in his 1986 book “Teachers and Machines.”

Dr. Cuban describes an “unrelenting cycle” that occurs with every new technology, whether radio, film, television or computers. First, reformers, who tended to be either executives or administrators, would promise to “revolutionize” the classroom with a new device. Money may be included; Seckinger received a grant for over $100,000 from Google.org, the tech company’s philanthropic arm, for A.I. research in 2024. That promise would be followed by breathless press around how any given technology would improve student learning, make the classroom more cost-effective and enhance teacher skills.

These guarantees always ignored all of the other pressures faced by teachers, who still had all the old responsibilities to balance while integrating new devices into their instruction. The next part of the cycle — after the glowing publicity and often uncritical imposition of the technology — is disappointment. Uptake by classroom teachers is inconsistent, often because the tech doesn’t work well. During my visit to Georgia, I observed a music teacher looking anxious as her eager fifth graders could not perform their block-coded melodies for me on their glitchy Padlet app.

Even if the devices work as advertised, they don’t instantly offer the productivity or test-score gains that were initially assured. The last part of the cycle that Dr. Cuban outlines is teacher bashing. Educators tend to be the ones blamed when the unrealistic expectations of administrators don’t materialize.

With A.I., we’re somewhere between Parts 2 and 3 of Dr. Cuban’s cycle, the point he describes as the “fickle romance” between technological innovation and the American classroom. We’re in the middle of a backlash against smartphones and Chromebooks in schools just as districts are starting to contend with the infiltration of A.I. New York City recently paused the creation of an A.I.-focused high school in Manhattan after protests from parents.

Even the founders of educational tech companies are starting to reconsider. In 2023, Sal Khan, the founder and chief executive of Khan Academy, introduced Khanmigo, an A.I. chatbot. “We’re at the cusp of using A.I. for probably the biggest positive transformation that education has ever seen. And the way we’re going to do that is by giving every student on the planet an artificially intelligent but amazing personal tutor,” Mr. Khan boasted in a TED Talk that year.

In 2026, Mr. Khan is softening some of those hyperbolic claims. “For a lot of students, it was a non-event,” Mr. Khan told Chalkbeat’s Matt Barnum in April. An Indiana teacher who was an early adopter of Khanmigo said many of her students found the bot frustrating, in part because it made mistakes, and in part because it didn’t often deepen students’ learning.

- Author: Jessica Grose, The New York Times