How AI Impacts Real Learning: From Performance to Lasting Understanding

Here’s a truth many educators hesitate to admit: students can ace every assignment and learn almost nothing, while those who struggle often gain the most. UCLA researchers Robert Bjork and Nicholas Soderstrom describe this paradox, highlighting the difference between performance—doing well in the moment—and learning, which creates lasting change in the brain.

“Performance is what looks good today; learning is what sticks for tomorrow.”

Performance may look impressive, like submitting a flawless essay created with AI. But true learning requires more than polished outputs. Barbara Oakley explains that mastery moves knowledge from conscious effort to automatic expertise. At first, tasks rely on declarative memory, demanding full attention. With practice, skills transition to procedural memory, allowing automatic execution. Learning to drive illustrates this: beginners focus on every movement, while experienced drivers navigate traffic effortlessly. When AI solves problems for students, these essential brain pathways may never form.

The Illusion of Understanding

Consider this scenario: students take a test immediately after a lesson and score 90%, but a week later, without reviewing, scores drop to 60%. High performance in the moment does not guarantee long-term memory retention.

“Using AI to write essays can make the work look perfect, yet students often skip the mental effort that leads to real understanding” paraphrased from Robert Pondiscio.

This is where productive struggle—the effort to retrieve and apply knowledge—truly builds learning. AI can improve short-term performance without fostering this struggle, creating the illusion of understanding.

AI: A Double-Edged Sword

AI can either accelerate learning or undermine it, depending on how it is used:

• Poorly implemented AI: A 2024 study in Turkey found students with unrestricted AI access performed 17% worse than peers without AI. ChatGPT-generated essays looked polished but reduced planning, self-evaluation, and cognitive engagement.

• Well-designed AI: Harvard and Stanford studies show that AI tutors can double learning efficiency when teachers scaffold tasks, ask probing questions, and maintain student thinking. In Nigeria, students achieved in six weeks what typically takes over a year and a half.

Preserving Cognitive Effort

The key to effective AI use is maintaining mental struggle. When AI does all the thinking, students never confront errors or reflect, removing opportunities for memory formation.

Teachers must design tasks and AI prompts that encourage problem-solving, reflection, and independent thinking.

Preparing Educators and Policymakers

AI alone cannot ensure learning. Schools must:

• Train teachers to integrate AI while preserving cognitive effort.

• Help leaders distinguish between polished outputs and real understanding.

• Equip policymakers with guidance, time, resources, and ongoing support.

• Build educator communities to share strategies and expertise.

In developing countries, foundational skills are often weaker. Students may accept AI outputs without questioning them, missing the struggle needed to develop expertise. Yet research shows that guided AI use can deliver remarkable learning gains, even for students with less prior knowledge.

The Path Forward

AI is neither inherently good nor bad. Its impact depends on how it amplifies human expertise rather than replacing it. The goal is not just producing flawless work, but cultivating thinkers who can apply knowledge creatively in new contexts.

When educators guide AI to support cognitive effort, students don’t just produce smart answers—they become smarter thinkers.

Learn More

• Learning How to Learn – Barbara Oakley’s Coursera course on memory, cognitive effort, and deep learning strategies.

• A Mind for Numbers – Practical techniques for mastering difficult subjects.

• Learning Versus Performance – Robert Bjork’s research on retrieval practice and desirable difficulties.

• The Learning Scientists – Evidence-based strategies like spaced practice and elaboration.

• UNESCO Science of Learning Portal – Global resources for educators.

Inspired by researchED Chile 2025 presentations by Barbara Oakley, Tom Bennett, Nidhi Sachdeva, Greg Ashman, and Rodrigo López.