By Rudy Malcom
Last fall, the University of California San Diego, one of the nation’s top public universities, reported that about one in eight incoming students demonstrated math proficiency below the high-school level — nearly 30 times the share recorded five years earlier.
The causes of that decline remain widely debated. But some worry that the rapid adoption of AI in schools could worsen the trend by allowing students to bypass the productive struggle needed to build basic competencies. Indeed, in a 2025 health advisory, the American Psychological Association (APA) warned that “students must not become overly dependent on AI, which could impede developing their own knowledge and skills.”
“How much of the fact that kids can’t do math has to do with the fact that they’ve let AI do their homework? I think it’s too soon to know the answer,” says Jonathan Zimmerman, a professor of the history of education at the University of Pennsylvania.
“It’s kind of a Wild West,” Zimmerman says of the use of AI in the classroom. “We started to adopt these tools without knowing whether they promote or inhibit learning and under what conditions.”
“We’ve invested in these tools, literally and figuratively, before we know what they do. And now, we’re in some ways incentivized to answer the question affirmatively: Of course they promote learning — because we bought them.”
AI tools used by students generally fall into three categories:
- Generative AI: ChatGPT, Google Gemini and Microsoft Copilot can support brainstorming, research and writing.
- AI tutoring and study tools: Tools like AskSia, Khanmigo and NotebookLM can help students review course materials providing explanations, answering questions and offering feedback.
- Writing and language tools: Grammarly and other AI writing assistants can help users improve their grammar, vocabulary and writing.
The APA cites research suggesting that AI’s organizational, summarization and problem-solving features can help students understand and retain key concepts. According to its advisory, “AI promotes growth through advanced questioning techniques that can stimulate critical thinking, scaffolding that provides step-by-step guidance and adaptive learning that delivers personalized feedback.”
Zimmerman, whose recent work examines controversial issues in public schools, believes there should be more research into the benefits and drawbacks of AI for students. The APA similarly called for “rigorous scientific investigation,” including longitudinal studies of the same individuals over an extended period, studies that identify causal relationships and long-term effects and studies of children from diverse communities.
The APA also stressed that understanding AI’s wide-ranging implications will require collaboration between a variety of stakeholders, such as scientists, ethicists and educators — along with parents, who should “model healthy behavior and critical thinking for adolescents,” the advisory reads.
It adds: “Parents and caregivers play a vital role in guiding and protecting adolescents as they navigate AI technologies.”
Common Sense Media offers these tips:
- Talk about the benefits and risks of using AI. For example, remind your child that AI tools are programs, not people.
- Try out AI tools together. This can help parents evaluate whether they are age-appropriate.
- Talk about biases in tech. Encourage your child to think critically about information provided by AI and to verify answers using trustworthy sources.
- Have conversations about plagiarism and cheating. Explain to your child that submitting AI-generated work as their own could violate academic integrity rules.
- Ask your child’s teachers and school for support.
Zimmerman is “glad to see people raising their voices” in Philadelphia’s Main Line suburbs, where some parents are mobilizing against the use of AI in classrooms and curricula (and against screens more broadly). “A certain amount of skepticism is hugely warranted,” he says.
Yet, the APA notes, parents often lack the time or capacity to learn about AI’s promises and pitfalls. Zimmerman encourages parents to ask, as the late social critic and educator Neil Postman did: To what problem is a new technology being framed as the solution? “What are we trying to correct for here? Is it actually correcting that problem? How do we know if it is? What other problems might be created along the way?”
Zimmerman, a former public school social studies teacher, also suggests that parents consider which tasks are best suited to AI tools. For instance, how can AI complement teachers rather than replace them? Is AI better at helping children learn, say, how to multiply or how to construct sentences — or vice versa?
More than half of teens (54%) surveyed by the Pew Research Center last fall said they used chatbots for help with schoolwork, with 10% reporting that they used AI for most or all of their assignments. More specifically, 40% said they used AI to research topics and solve math problems, while 35% said they used it to edit their writing.
About half of respondents (51%) said chatbots were at least somewhat helpful for completing schoolwork, while only 3% said they were of little to no help.
Zimmerman says that discussions about AI should also account for the social side of learning. “Schooling is not just a place where we learn algebra or our times table — it’s a place where we learn how to cooperate, how to converse and how to disagree,” he says. “These are all really important social and, indeed, civic functions.”
These functions include learning the value of integrity — an area where AI may pose another challenge. More than half (59%) of teens surveyed by Pew said using AI to cheat happens at least somewhat often at school, with about a third (34%) saying it happens extremely or very often. Some research has linked academic cheating to dishonesty later in life, though the long-term effects of students’ AI use on academic integrity and broader ethical behavior remain unclear.
In the 1920s, Thomas Edison predicted, incorrectly, that motion pictures would replace textbooks within two years — a reminder, Zimmerman says, that sweeping claims about new educational technologies have a long history. “None of us know how AI can change this landscape,” he says, “but we should be guarded about these techno-futuristic predictions that it’s going to change everything.”
In their 1995 book “Tinkering Toward Utopia,” David Tyack and Larry Cuban argue that many aspects of schooling have proven resistant to reform — a concept the historians call the “grammar of schooling.” Despite decades of technological and social change, many basic structures of schools have remained largely intact. For example, schools largely still follow the same daily schedule, opening around 8 a.m. and closing around 3 p.m.
“The grammar of schooling — what schools should look and feel like — is so deeply inscribed in our bones, and what that means is that it changes very slowly,” Zimmerman says. “I’m a historian. I study the past. I can’t predict the future. But if the past is any guide at all, we may discover that AI changes education less than we thought it would.”
