About 40 percent of ChatGPT users are under the age of 24. Hyoung Chang/The Denver Post
With ChatGPT now widely available to students around the globe, concerns about cheating and the erosion of critical thinking skills are growing. OpenAI is well aware of these issues and is working to address them. Speaking today (July 9) at the AI for Good Summit in Geneva, Kevin Mills, OpenAI’s head of education and government, outlined the company’s efforts to explore how A.I. tools can be used to support, rather than replace, meaningful learning.
“We know that true learning takes friction. It takes struggle,” said Mills. “You have to engage with the materials, and if students offload all of that work to a tool like ChatGPT, they will not learn those skills and they will not gain that critical thinking. That said, when ChatGPT is used correctly as a learning assistant and as a tutor, the results are powerful.”
Though ChatGPT has only been publicly available for less than three years, OpenAI has already partnered with the government of Estonia to introduce customized generative A.I. tools into public schools in the European nation. The company has also launched collaborations with Harvard University, Arizona State University and campuses across the California State University system.
Given that 40 percent of ChatGPT users are under the age of 24—and that learning is the platform’s number one use case, according to Mills—the need to fine-tune guardrails is becoming increasingly urgent. Pew Research reports that twice as many teens now use ChatGPT for schoolwork compared to 2023, with nearly one-third of teen respondents saying it’s acceptable to use the tool to solve math problems.
In response, Mills said OpenAI is actively researching what appropriate A.I. use in education looks like, with plans to share that guidance widely and rapidly with educators around the world.
As part of this initiative, OpenAI yesterday announced a new partnership with the American Federation of Teachers to launch the National Academy for AI Instruction, which aims to equip more than 400,000 educators with practical A.I. fluency by 2030.
The company also recently launched OpenAI Academy, offering “many online free courses on A.I. literacy and skill building for all sorts of personas,” said Mills.
While OpenAI’s research on how to most effectively integrate A.I. into teaching and learning is still underway, Mills emphasized that thorough longitudinal studies are on the horizon. But these take time. In the meantime, educators are turning to existing studies for guidance. One such example: A.I. in education “can help students acquire problem-solving and online collaboration skills, enhancing learning quality,” wrote the University of Saskatchewan’s Shan Wang in a 2024 academic paper.
However, play-based, hands-on and social learning remain critical components of K–12 education, according to the Center for American Progress. “It’s important to think critically about whether adoption of A.I. tools will further divorce students from how their brains are primed to learn through modalities such as movement, play and social interaction,” the research and advocacy organization wrote in a 2024 report.
OpenAI has begun to consider the social dimension of learning through initiatives like the ChatGPT Lab, where “students come together to share how they’re using ChatGPT to further their education,” said Mills. This kind of peer-to-peer knowledge sharing is key, he added: “One of the coolest things about A.I. technology is that so much of the innovation actually happens after the technology has left the lab.”