MIAMI» One morning in April, Tracy Lowd, a social studies teacher in Miami, tried a new approach to making government policy come alive for her high school students. She used artificial intelligence chatbots to role-play American presidents.
Her class at Southwest Miami Senior High School had already read about John F. Kennedy and discussed his campaign for “new frontier” economic and social policies. Now Lowd asked two dozen 11th graders to open their laptops and type a prompt into Google’s Gemini chatbot: “Act like President Kennedy. What was the new frontier?”
The chatbot quickly spat out paragraphs of Kennedyesque text, including phrases like “my fellow Americans.”
Then Lowd asked her students to analyze whether the chatbot simulations accurately reflected the Kennedy speeches they had studied. The teenagers’ verdict: The simulations were “awkward,” “weird” and yet still credible.
“It did a very good job of impersonating JFK,” said Ashley Acedo, 17.
Miami-Dade County Public Schools, the nation’s third-largest school district, is at the forefront of a fast-moving national experiment to embed generative AI technologies into teaching and learning. Over the last year, the district has trained more than 1,000 educators on new AI tools and is now introducing Google ch atbots for more than 105,000 high schoolers — the largest U.S. school district deployment of its kind to date.
It is a sharp turnabout from two years ago, when districts like Miami blocked AI chatbots over fears of mass cheating and misinformation. The chatbots, which are trained on databases of texts, can quickly generate humanish emails, class quizzes and lesson plans. They also make stuff up, which could mislead students.
Now some formerly wary schools are introducing generative AI tools with the idea of helping students prepare for evolving job demands. Miami school leaders say they also want students to learn how to critically assess new AI tools and use them responsibly.
“Every student should have some level of introduction to AI because it’s going to impact all of our lives, one way or another, in the tools we are using in our jobs,” said Roberto J. Alonso, a Miami-Dade school board member.
The AI about-face in schools comes as President Donald Trump and Silicon Valley leaders are pushing to get the technologies into more classrooms.
Some tech billionaires are promoting grandiose visions of the AI systems as powerful tutoring bots that will instantly tailor content to each student’s learning level. Google and OpenAI, the maker of ChatGPT, are fiercely competing to woo education leaders and capture classrooms with their AI tools.
Industry giants like Microsoft argue that training young Americans in workplace AI skills has become a national economic necessity to compete with China. Last month, Trump agreed, signing an executive order intended to spur schools to “integrate the fundamentals of AI in all subject areas” and for students “from kindergarten through 12th grade.”
If the classroom AI crusade succeeds, it could remake teaching and learning, in part by casting chatbots as the intermediaries that students turn to first for tutoring and feedback — before teachers see their work.
The AI gambit could also end up eroding important skills like critical thinking, researchers say, or lead students to over-rely on chatbots.
A recent report from RAND, a research organization, said, “The likelihood that generative AI tools are leading to measurable improvements in teaching and learning is low.”
The classroom chatbots could also lose their luster when the next tech innovation comes along. A previous Silicon Valley campaign for computer science in schools now faces an existential crisis, with funders shifting their attention to AI literacy.
Miami-Dade County is joining the classroom AI wave by embracing the new tools as practical widgets that teachers and students can use — albeit with a critical lens and frequent fact-checking.
“AI is just another tool in the arsenal of education,” said Daniel Mateo, the assistant superintendent of innovation at Miami- Dade schools and the architect of the district’s AI initiative. As with other instructional tools, he said, “we have to make sure that we use it ethically, that we use it responsibly, that we have certain guardrails in place — and that all happens through our vetting process.”
The Miami effort is part of a larger push to spread AI tools and literacy in Florida classrooms. Last year, the University of Florida set up a statewide education task force — which includes more than two dozen districts, among them Miami-Dade County, Broward County and Palm Beach County schools — to develop AI guidelines for local schools.
“AI is already coming into schools, and so not having an informed, strategic approach to considering AI is really risky,” said Maya Israel, an associate professor of computer science education at the University of Florida overseeing the group.
In 2023, Miami-Dade schools initially blocked AI chatbots such as ChatGPT. Then Mateo, a tech enthusiast who has “smarted out” his house with robot vacuums and automated thermostats, began to consider school uses for the new AI tools. He envisioned chatbots helpfully summarizing reports for principals and suggesting new lesson ideas for educators.
If the district trained teachers on the systems, Mateo figured, Miami educators could then help students use chatbots for learning, not cheating.
Technology staff at Miami- Dade schools first spent months assessing nearly a dozen different AI tools for accuracy, privacy and fairness. The top contenders: Google’s Gemini, OpenAI’s Chat- GPT and Copilot from Microsoft.
Members of Mateo’s team, posing as teenage hackers, also entered rude comments to see if they could prompt the chatbots to produce racist, violent or sexually explicit responses.
“We were tasked with trying to break AI,” Jeannette Tejeda, a district instructional technology specialist, explained. “We asked the AI the most inappropriate questions you can imagine.”
The district ultimately chose Gemini for its students, partly because Google offered certain content and privacy guardrails for teenagers — including not using information that students entered into the chatbot to train the company’s AI models, Mateo said.
Next, the school district developed AI training workshops for its 17,000 teachers.
Called “the AI Institute,” the program now offers dozens of live virtual sessions for teachers. Course descriptions include: “Transform your lesson planning with AI!” and “Discover how AI language models can revolutionize teaching writing.”
This spring, local high schools tested the chatbot with students. One of the first was Southwest Miami Senior High, a sprawling concrete complex with royal purple trim, with about 2,500 students. The tech-friendly school offers a variety of Advanced Placement and computing classes, including AI Fundamentals, a collegelevel course developed by the University of Florida.
Maria Chirino, an English teacher at the high school, recently gave the 10th graders in her language arts class a writing assignment. The class had been reading “Oedipus Rex,” the Greek tragedy. So Chirino asked the teenagers to write a paragraph on whether humans have free will to shape their own futures or are controlled by fate.
Only instead of grading her students’ writing herself, Chirino for the first time assigned her class to ask the Google chatbot for feedback.
“I was initially skeptical because I thought kids would try to generate their paragraphs using the AI, instead of writing their own,” Chirino said. Then she tried the chatbot exercise with her 12th-grade literature class and found students liked the bot’s instant feedback.
“They could rewrite the paragraphs right away,” Chirino said, “instead of having to wait a day or two before they would get their essays back from me.”
Among the 10th graders, Karen Valdeon, 16, argued for free will, writing that humans “make different choices that can alter the outcomes” of their lives. Then she pasted her paragraph into Gemini. She also put in Chirino’s grading standards, which told the chatbot to give students points for making clear arguments or effectively using examples to back up their claims.
“Your ideas flow pretty well,” the chatbot said. Noting that Valdeon’s thesis statement was clear and her examples were relevant, Gemini awarded her 5 points, a perfect score. “I didn’t notice any major errors in your writing,” it added.
Then Chirino asked the students to submit their paragraphs directly to her for grading. She said she would also review the chatbot’s assessments of students’ writing.
Mateo said he hoped the district’s chatbot rollout would ultimately transform learning, by quickly providing useful information to students who might, say, need immediate help with a calculus problem late at night.
Many teenagers are already using AI chatbots outside school, said Jorge M. Bulnes, the principal of Southwest Miami Senior High.
“We have an obligation to help them navigate that usage,” he said. Back in social studies, Lowd said she had noticed some hiccups, like when Gemini missed an entire portion of a writing assignment. She now uses such errors to teach students to watch out for AI-generated mistakes.
After her class had finished analyzing the chatbot impersonations of Kennedy, she asked the teenagers to close their laptops. Then she assigned an essay — to be written in longhand — comparing different presidential policies.
“I’m teaching students to use AI as a tool that’s helpful, like a book or a dictionary,” Lowd said. “I’m not telling them to use it to get the answer.”