The Voices section is a place for physicians, staff and community leaders to share their perspectives on all things healthcare. Ashley Phillips, PhD is a licensed pediatric psychologist at Valley Children’s and a researcher at the intersection of education, psychology and neurodevelopment.
MADERA, Calif. – Artificial intelligence is becoming more common in schools, with many districts exploring how it can support teaching and learning. As AI use grows, educators face an important question: Should students use artificial intelligence as part of the learning process?
Research on artificial intelligence in education is still limited, while cognitive science research shows that students learn best when they are actively engaged, work through challenges, collaborate with others and receive meaningful feedback.
Current AI tools may weaken these learning conditions by encouraging students to rely on technology instead of thinking independently, solving problems or interacting with teachers and classmates.
Because these skills are critical to student development and there is little long-term evidence about AI’s effects on learning, school districts should take a cautious approach to student AI use.
Key Findings
Students using AI tools become passive learners
- Participants who use large language models (LLMs) put forth less effort, feel less ownership of their work, learn fewer new concepts and produce less complete responses (Melumad & Yun, 2025).
AI tools create a false sense of learning
- Bastani et al. (2024) found that students who used GPT scored 17% lower on a math test than non-AI users when the AI tool was removed.
- AI use can cause people to overestimate what they know and create a false sense of mastery (Matueny & Nyamai, 2025).
Learning has a social component that AI cannot replace
- Generative AI can reduce student-to-student interaction, weakening learning communities (Hou et al., 2025).
- AI can create a “zone of no development” when ongoing assistance replaces productive struggle and support from teachers and peers that helps students learn (dos Santos & Birdwell, 2025).
There are equity concerns with AI use
- Twenty percent of teens in households earning less than $30,000 reported doing most of their homework with AI, compared with 7% of teens in households earning more than $75,000 (Pew Research Center, 2026).
Generative AI, particularly large language models, is a new technology with a limited research base in K-12 education. Some AI applications may be useful for teachers, but schools do not need to wait for AI-specific studies to understand potential risks.
Decades of cognitive science research show that long-term learning depends on retrieval practice, productive struggle and support from teachers and peers within a student’s zone of proximal development, the space between what a student can do independently and what a student can do with guidance (Dehaene, 2020).
Current evidence suggests that student AI use can lead to shallower learning, a false sense of understanding and weaker peer relationships (Melumad & Yun, 2025; Bastani et al., 2024; Hou et al., 2025). These risks may be greater for lower-income students (Pew Research Center, 2025; Brookings, 2026).
Education policy should be guided by the science of how children learn, not by assumptions about technological progress.
Policy Implications
Delay widespread adoption
- Require pilot program data that shows measurable improvements in student learning before districtwide implementation.
- Apply the same evidence standards used for other educational interventions.
Restrict AI use in foundational skill development
- Limit AI use in reading, writing and mathematics instruction. These foundational skills require the mental effort that builds lasting knowledge and competence.
Treat AI as a potentially high-risk educational intervention
- Fund and require peer-reviewed research on long-term academic, cognitive and equity outcomes.
- Conduct equity impact assessments before implementation.
- Teach students about the risks of relying on AI instead of developing their own thinking skills.
- Develop AI literacy programs that promote age-appropriate and responsible use.
Prioritize investments in teacher-led instruction
- Focus AI investment on teacher-facing tools, such as lesson planning, rubric creation and data analysis, which offer the greatest benefits with the lowest risk.
- Provide educators with professional development in AI before expanding student access.
- Teachers should remain the instructional experts and final decision-makers. AI should support teachers’ work, not replace professional judgment.
References
Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, Ö., & Mariman, R. (2024). Generative AI without guardrails can harm learning: Evidence from high school mathematics. Proceedings of the National Academy of Sciences. https://www.pnas.org/doi/10.1073/pnas.2422633122
Dehaene, S. (2020). How we learn: Why brains learn better than any machine . . . for now. Viking.
dos Santos, E. C., Jr., & Birdwell, T. (2025). The unspoken crisis of learning: The surging zone of no development. arXiv. https://arxiv.org/abs/2511.12822
Hou, I., Man, O., Hamilton, K., Muthusekaran, S., Johnykutty, J., Zadeh, L., & MacNeil, S. (2025). “All roads lead to ChatGPT”: How generative AI is eroding social interactions and student learning communities. In Proceedings of the 30th ACM Conference on Innovation and Technology in Computer Science Education (Vol. 1). ACM. https://dl.acm.org/doi/10.1145/3724363.3729024
Matueny, R. M., & Nyamai, J. J. (2025). Illusion of competence and skill degradation in artificial intelligence dependency among users. International Journal of Research and Scientific Innovation, 12(5), 1725–1738.
Melumad, S., & Yun, J. H. (2025). Experimental evidence of the effects of large language models versus web search on depth of learning. PNAS Nexus, 4(10), pgaf316. https://academic.oup.com/pnasnexus/article/4/10/pgaf316/8303888
Pew Research Center. (2026, February 24). How teens use and view AI. https://www.pewresearch.org/internet/2026/02/24/how-teens-use-and-view-ai/
Winthrop, R., Burns, M., Luther, N., Venetis, E., & Karim, R. (2026). A new direction for students in an AI world: Prosper, prepare, protect. Brookings Institution. https://www.brookings.edu/articles/a-new-direction-for-students-in-an-ai-world-prosper-prepare-protect/
