U.S. public schools are increasingly adopting a strategy of incorporating artificial intelligence into the classroom, moving away from outright bans to teach students about the technology's limitations, including its tendency to generate inaccurate information, a phenomenon known as "hallucination."
This approach to AI literacy aims to equip students with skills for an AI-driven future while preventing over-reliance on chatbots for their thinking. The concept of AI literacy is evolving, with a growing consensus among educators that it involves more than just learning to use generative AI tools or craft effective prompts. Rebecca Winthrop, director of the Center for Universal Education at the Brookings Institution, emphasizes that "good AI literacy includes knowing when not to use it."
As the new school year begins, many districts are implementing training programs. In Charleston, South Carolina, educators participated in sessions demonstrating AI's flaws, such as generating a world map with numerous factual and spelling errors. Amanda Bickerstaff, founder and CEO of AI For Education, used this demonstration to highlight how students can learn to critically evaluate AI outputs.
Nationwide, studies indicate that a majority of teenagers and teachers are using AI for schoolwork, often without formal training or a deep technical understanding of how it functions. This mirrors concerns that arose with social media, where a lack of guidance led to issues related to addictive algorithms and mental health impacts. Schools and parents are eager to avoid similar pitfalls with AI.
Charleston County School District, the second-largest in South Carolina, developed an AI policy after consulting with teachers, students, and parents. Student focus groups revealed widespread, unguided AI use, prompting the district to implement Phase 2 of its plan: comprehensive training for teachers and students on AI functionality and effective, ethical use.
The student curriculum includes scenarios showing how generative AI might present fabricated studies or inaccurate information with authority, stressing the importance of verifying all factual data obtained from these tools. Data privacy is also a key component, cautioning students against sharing personal information due to the potential for conversation storage, data training, and leaks.
Ray Knauer, a high school teacher, found the training provided "teachable moments" for his AP Research class, planning to use examples of AI bias and hallucinations to foster student discussions on critical analysis and deeper inquiry beyond immediate AI-generated answers.
Utah has taken a proactive statewide approach. In 2024, the state appointed Matt Winters as its AI education specialist, the first such full-time position in the U.S. Winters has led training for over 7,000 teachers and is assisting districts in developing AI policies, which are mandated by state law by July 2027. Chris Agnew, director of Stanford University's Generative AI for Education Hub, noted Utah's coherent, top-down method as beneficial for managing significant change.
Other states, including Maine, West Virginia, and Georgia, have followed Utah's model. However, a fragmented approach persists in many states, raising concerns about equitable access to AI literacy training, particularly for students in under-resourced districts. The investment required for extensive training can exacerbate these disparities.
Utah's training covers AI fundamentals, bias, hallucinations, ethical considerations, and data privacy. Emma Moss, who oversees AI at Canyons School District, stressed the importance of accessible training, with her district offering sessions statewide, funded by a significant grant. The state has also facilitated access to AI tools through negotiated data privacy agreements and discounted pricing.
Winters advises against blanket bans, stating that "AI literacy is much larger than saying: ‘Don’t use AI.’ It’s about how we can prepare kids for the future." Kristina Yamada, the state Board of Education's digital technology specialist, highlights the need to shift educators' mindsets regarding AI and cheating. She emphasizes that teachers must engage students in conversations about the fallibility of AI, similar to how they discuss human deception, and encourage them to develop problem-solving skills when AI-generated solutions fail.