When the public talks about AI in the classroom, it’s often to fret over students who get “chat” to write their essays. But the AI revolution in education is happening on an institutional scale—and it goes much deeper than homework shortcuts.
Tech companies are releasing lineups of AI-focused education devices, and school districts are buying them in bulk. Educators are reconfiguring their pedagogical practices, departments their curricula. When applied effectively, AI teaching tools have helped K–12 teachers streamline workflows and personalize lesson plans to accommodate diverse learners. In higher education, faculty are seeing unprecedented course completion rates on AI-powered instruction platforms.
Yet there’s plenty of evidence that AI use weakens students’ critical thinking and research skills. Many educators believe that AI also dilutes human connection and the value of in-person learning, which has only just been revived post COVID-19.
So what’s the solution?
A Handbook for “Healthy Skeptics”
According to learning sciences faculty in the Adelphi University Ruth S. Ammon College of Education and Health Sciences, there’s no one right answer.
“The most empowering thing we can do for educators right now is help them see that they don’t have to demonize or fetishize AI,” said Elizabeth De Freitas, PhD, professor of learning sciences. But educators should also slow down and reflect. “Rather than simply leaping to a particular use of AI, we have to spend more time understanding the nature of our use and how it comes to us historically.”
Dr. De Freitas and her Adelphi colleague Matthew Curinga, EdD, associate professor of learning sciences, co-edited Learning Under Algorithmic Conditions (University of Minnesota Press, 2026), a new collection of 27 essays on machine learning from scholars in media studies, philosophy, computer science, math, linguistics and education. The book serves as an interdisciplinary guide for educators trying to make sense of how human learning is being newly reconceived by AI.
“There’s a lot of polarization in the AI conversation. You’re either for or against it,” Dr. Curinga said. “But the book is not ‘do this, don’t do that.’ We’re providing the nuance that educators need to evaluate AI for themselves, to develop their own theories and philosophies, to become healthy skeptics of technology.”
Beyond the Hype and the Headlines
To Dr. De Freitas, AI is simply another mode of literary production. Much like a human journalist might speak to sources and then draft a story, AI collects data from the internet and then generates hypothetical claims. And, much like a human journalist, AI has a subjective perspective. It’s not a truth teller or sage, she said, nor is it a “black box”—a phenomenon that’s beyond the ken of anyone but computer scientists. “There are so many ways to hook into AI on so many different levels.”
AI isn’t all-powerful, either. “We can’t just read the headlines that tell us AI will replace a million jobs by 2050 and shrug our shoulders,” Dr. Curinga said. He and Dr. De Freitas reject the concept of technological determinism, in which AI is an inevitable force driving its own future, impervious to human intervention. Instead, their work equips educators with the agency to keep learning firmly human-driven, even when it involves machines.
Research That Resists Easy Answers
Other learning sciences faculty are resisting the good-bad binary in their research and teaching.
Aaron Chia Yuan Hung, EdD, associate professor of education, who’s in the early stages of research on generative AI and learning, said his interest is “less on whether generative AI helps or hurts, but what kind of pedagogy has to be in place for it to help a specific group of learners.”
So far, Dr. Hung’s findings have been inconclusive. “There’s compelling evidence that heavy reliance on large language models carries a real cognitive cost,” he said, including effects that lingered even after students stopped using the technology. But other studies have shown positive results: an increase in creativity when brainstorming, an AI-powered tutor that’s engaging students more deeply than active instruction. “Put together, these studies suggest that GenAI used as a substitute for effort has real costs, and that GenAI used as a well-designed instructional or creative tool can offer meaningful support.” What a user is actually asking the tool to do, Dr. Hung says, is likely the most important variable.
Tracy Hogan, PhD, professor of learning sciences, is intrigued by the gendered dimensions of AI technology. Her research investigates the relationship between a user’s background and how they interact with chatbots—specifically, if users with computer science experience are representing chatbots through a neutral gender lens, while less well-informed users are representing their bots through a specific gender identity. She’s brought these ideas into her Critical Literacy in Mathematics and Science Education class, where she works with students to explore inherent biases in AI input and output.
Training Tomorrow’s Educators for the AI Classroom
Whether producing their own research or referring to others’, Adelphi faculty are applying evidence-based teaching methods in their classes, preparing a generation of educators to adapt to an ever-evolving AI landscape.
In the University’s Manhattan Center, students in the PhD in Learning Sciences program engage in complex machine learning scenarios in both the classroom and the center’s STEAM Innovation Lab. Though the STEAM lab is largely analog, Dr. Hogan encourages her students to play around with AI tools when designing and coding projects. Most of what these tools generate is incorrect or incomplete, so students are forced to draw on their own knowledge to critically examine AI’s output—a process that, in the end, reflects the STEAM program’s commitment to learning by experience and experimentation.
In Dr. Hung’s classroom, students are able to peek inside the “black box” of AI. Many of his students don’t actually know how it works, even though they use it regularly. For Dr. Hung, explaining the basics—”why GenAI responses aren’t like a search engine’s, why it’s prone to hallucination, why it’s suited to some tasks and not others, why it almost never says it doesn’t have an answer”—is easy enough. The real challenge is keeping up. “More than once in my first semester teaching this class, a model we’d just discussed got a significant update midsemester that changed the conversation,” he said.
Although AI tools are imperfect, Adelphi learning sciences faculty aren’t opting out altogether. There’s much good that can be done with these tools, they believe, if applied rigorously and ethically. Most importantly, today’s students must be ready to lead the classrooms of tomorrow.
As Dr. Hung puts it, “Our students’ future students won’t opt out of AI, and more workplaces will expect graduates to arrive with some baseline AI literacy. The risks are real enough to design around carefully, but the technology isn’t going anywhere, so pretending otherwise doesn’t serve our students.”