Key points:
Recent findings on the negative impacts of AI on learning might be sparking national debate, but they are unsurprising to learning scientists. In fact, these results highlight a long-standing U.S. trend of using what “feels right” or “sounds good” instead of following well-established education research.
A recent Massachusetts Institute of Technology study shows that students’ over-reliance on technology in general, and AI in particular, bypasses essential learning processes during crucial stages of childhood and adolescent cognitive development. Examples of technology interfering with the development of critical thinking skills are abundant. They include replacing handwriting with keyboarding, reducing the importance of students’ automatic recall of foundational knowledge, and giving students answers before they engage in productive struggle.
AI should undoubtedly play a role in supporting learning, but it must be in a way that enhances, not interferes with, core learning science principles. Responsible and effective AI use requires strong data and oversight. When AI is grounded in a complete, accurate view of each student, educators can quickly make instruction more contextually relevant and deliver practice within each student’s zone of proximal development. Used this way, AI becomes a tool that deepens thinking, supports personalization, and accelerates meaningful academic growth.
However, to see the results of our failure to use technology in a way that follows, not circumvents, evidence-based guidance, one need only look at our students’ declining test scores. Not only are scores lower absolutely, but also relatively to our Asian and European counterparts who have wisely been managing technology usage, particularly for younger students.
What can happen when we embrace learning science
A notable exception to our collective national dissonance with the learning sciences is the widespread adoption of the science of reading requirements in more than 40 states since 2019. While this recent success provides a beacon of hope, the full history is far more complex and leaves much uncertainty about how schools respond to AI.
The “Mississippi Miracle” started the science of reading movement when Mississippi essentially went from worst in the nation to top 10 in NAEP fourth-grade reading scores in just six years. What’s less well-known is that the Mississippi adoption came 20 years after the National Reading Panel report left little debate about the best way to teach students to read.
Even then, Mississippi’s spectacular performance wasn’t enough. Rather, it was the groundswell of outrage from parents based on their firsthand experiences during the pandemic and spurred on by the 2022 podcast, Sold a Story, that led to the near nationwide mandate for evidence-based reading practices.
It’s unclear what spark could ignite a national mandate around AI and learning science. It might be family pushback against the $30 billion market for devices in schools. Or professional health advisories about AI and adolescent well-being.
The workers behind the ‘miracle’
To be clear, the Mississippi Miracle was no miracle. It came about through courageous leaders willing to put aside wishful thinking about technology and instead embrace the science–and associated hard work–of making systemic changes to properly teach kids how to read.
Glimmers of this courageousness are shining from organizations that lift up the most essential elements of effective learning, address ethical considerations around AI use, and highlight the complexity of human thought, which integrates emotion, context, nuance, and embodied experience. For instance, the Collaborative for Academic, Social, and Emotional Learning recently dedicated several conference sessions to the connections between social-emotional learning and AI.
At the state and local levels, Mississippi legislators and education leaders did the boots-on-the-ground work. They changed literacy policies, implemented comprehensive strategies, adopted new standards, hired additional literacy coaches, and spent years honing communications and convincing families and educators to give the science-based approach time to demonstrate impact.
The real question now is not what works in education. The science of learning has already answered that. The question is whether we have the collective will to ensure AI in schools is guided by that same evidence–and fueled by the kind of complete, high-quality student data that allows it to truly support learning.
Strong AI will only come from strong data, grounded in learning science and used with intention. Without it, we risk repeating the very mistakes we are trying to solve.
