Learning Still Takes Effort: Why Easy Is Not the Same as Learned
If a tool makes the work feel smooth, it is tempting to conclude that learning has happened. The feeling is often wrong.
When practice feels easy, it is tempting to conclude that learning has happened. Decades of research, and a growing set of AI studies, suggest we should be more careful.
Elizabeth and Robert Bjork have long argued for desirable difficulties: spacing practice over time rather than cramming, mixing problem types rather than blocking them, and using tests as learning events. These conditions can slow performance in the moment. They also tend to support later recall and transfer. The word desirable matters. A difficulty is useful only when the learner has enough knowledge and support to work through it. Frustration without a way forward is not a teaching strategy.
Dunlosky and colleagues (2013) found a similar pattern. Practice testing and distributed practice had relatively high utility. Rereading and highlighting, which often feel productive, had much lower utility. The techniques that work best ask students to retrieve, generate, and come back to material over time, and to teach it in their own words. They ask for effort.
The techniques that work best ask students to retrieve, generate, and come back to material over time, and to teach it in their own words.
What Feels Productive, What Works
Lower utility
Often feels productive
- Rereading
- Highlighting
Higher utility
Asks for effort
- Practice testing
- Distributed practice
- Immediate correction
- Teaching to a grown-up
- Presenting to a small group
A student closes the book and tells you what happened, or writes down what they remember from a chapter, then checks the gaps. That discomfort is part of the practice. A tool that supplies the retelling too quickly can remove the retrieval that would have strengthened memory.
That is also a fairer test than a page that looks complete. If the student can tell you the story, or explain a problem the next day, without the tool, something has likely been learned. If they can only recognize the answer when it is in front of them, the fluency was a feeling, not a skill.
If a child can only recognize the answer when it’s in front of them, they haven’t learned it yet. They’ve learned how it feels to see it.
Feeling of Learning Is Not a Reliable Meter
Students and adults can misread that effort. In a 2019 study, Deslauriers and colleagues compared a polished lecture with a more effortful active-learning lesson in undergraduate physics. Students learned more from the active class. They also reported feeling that they had learned less. The smoother lesson felt better. It taught less.
Felt Versus Learned
Students in a polished lecture felt they learned more. Students in a harder, hands-on class actually did. How learning feels is a poor guide to how much happened.
That finding is not an AI study, and it should not be stretched into a claim about every classroom. It is still a warning about intuition. When work feels fluent, we may overestimate what has been learned. Generative AI can intensify that mismatch, because it is so good at producing fluency. Cramming a study guide the night before can feel great. Spacing review over two weeks is usually stronger.
A growing body of AI-specific evidence points in the same direction, with important limits. In one randomized study with Year 10 students in England, Kreijkes and colleagues found that note-taking—alone or combined with a chatbot—supported comprehension and retention better than using the chatbot alone. Students preferred the chatbot and thought it was more helpful, even though it led to the weakest learning. Stanford’s SCALE Initiative, in a November 2026 review, reached a broader conclusion: AI tools can improve students’ learning experiences even when they do not improve learning outcomes. Positive feelings are not a substitute for later independent performance. Parents and teachers are not wrong to care about motivation. Students are more likely to return to work that does not feel punishing. The mistake is using comfort as the only meter. A lesson can be encouraging and still ask for retrieval, spacing, and a first unassisted attempt. The aim is not misery. It is a later explanation the student can actually give.
Teenagers in a recent study preferred studying with a chatbot and rated it the most helpful. It produced the weakest learning. Judge a study tool by what kids can do afterward, not how it felt.
Effort First, Then Help
This is not an argument for making childhood grim. Children need encouragement, clear explanations, and tasks they can complete. The point is narrower. If the goal is learning, someone has to do the cognitive work: retrieving a word, attempting a solution, putting an idea into a sentence. The sequence matters: effort first, help second, then a check on what the student can do without the help.
Before a child opens a chatbot, have them try the question on their own. Effort first, help second, then a check with the tool closed.
Before notes or a chatbot enter the picture, have the student attempt the question. Compare only after that first attempt exists. Mix a few problems from this week with a few from last month. Try a handful of items again the next day, not only on the night they were new. Homework is a common place for this mismatch. Finishing a set with a chatbot can take ten minutes and feel efficient. Attempting the first items alone, then using help only on a stuck step, takes longer. It also leaves a record of what the student could do. Tomorrow’s check—one mixed problem, or a short explanation with the tool closed—tells you more than last night’s completed page. When work looks easy, ask whether the student did the remembering, the writing, or the solving—and whether they can explain it without the tool.
What Shows Learning Has Not Changed
The tools are new. The evidence of learning is not. Long before chatbots, a teacher knew a student had learned something when the student could recall it without notes, use it on a problem they had not seen before, explain it in their own words, or teach it to someone else. That is still the test. AI changes how quickly a finished page can appear. It does not change what a student has to be able to do once the page is put away. A tool can help a child practice. It cannot do the remembering for them.
The tools are new. The evidence of learning is not: recall it without notes, explain it in your own words, teach it to someone else.
How Parents and Educators Can Approach This
Easy, fluent work can feel like learning and still leave knowledge thin. Keep some effort inside the week—retrieval, spacing, a first unassisted attempt—then offer help. The test is the one it has always been: whether the student can explain the idea later, without the tool.
When AI can finish any page, a finished page stops being proof of learning. Ask your student to explain it the next day, with the tool closed.
As a parent
Ask your student to teach you something they are learning at school, complete with diagrams and new vocabulary. Just play the role of an interested learner and ask questions.
Working one-to-one
After a session that felt smooth, close the notes or the chat and ask for one mixed or delayed item. If the student squirms, say that the discomfort of not quite remembering is part of the practice, not a sign the method failed. Ask the student to create their own questions and rate whether each is easy or hard to answer. Then probe further: ask for tricky questions, the kind that show they understand the subtleties.
In the classroom
You cannot make every minute effortful, and you should not. Work in spaced retrieval, more open-ended test questions, and class discussions of tricky questions and the tricky parts of a topic.
The Research
- Bjork, E. L., & Bjork, R. A. (2011). Making things hard on yourself, but in a good way: Creating desirable difficulties to enhance learning. In M. A. Gernsbacher et al. (Eds.), Psychology and the Real World (pp. 56–64). Worth Publishers. bjorklab.psych.ucla.edu/wp-content/uploads/sites/13/2016/04/EBjork_RBjork_2011.pdf
- Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J., & Willingham, D. T. (2013). Improving students’ learning with effective learning techniques. Psychological Science in the Public Interest, 14(1), 4–58. doi.org/10.1177/1529100612453266
- Deslauriers, L., McCarty, L. S., Miller, K., Callaghan, K., & Kestin, G. (2019). Measuring actual learning versus feeling of learning in response to being actively engaged in the classroom. Proceedings of the National Academy of Sciences, 116(39), 19251–19257. doi.org/10.1073/pnas.1821936116
- Kreijkes, P., Kewenig, V., Kuvalja, M., Lee, M., Hofman, J. M., Vitello, S., et al. (2026). Effects of LLM use and note-taking on reading comprehension and memory: A randomised experiment in secondary schools. Computers & Education, 243, 105514. doi.org/10.1016/j.compedu.2025.105514
- Walker, S., Fesler, L., Martinez Claeys, J. P., Agnew, C., & Loeb, S. (2026). The evidence base on AI in K-12: A November 2026 review. Stanford SCALE Initiative. scale.stanford.edu/ai/evidence-base-review/evidence-base-ai-k12-november-review
More Posts
When Help Becomes a Crutch: AI Tutoring and Overreliance
An AI tutor can raise a child’s practice scores and still leave them weaker without it. Effective help leaves the student equipped to do the next problem unassisted.
Nov 18, 2026 · 4 min readAI and LearningThe Foundations That Still Matter: Literacy and Numeracy
Most young children do not yet have their own AI tools. The work now is to build reading, number sense, and the habit of explaining what they know, so they are literate and discerning before that access arrives.
Nov 17, 2026 · 6 min readAI and LearningWhat Should Kids Still Learn When AI Can Give Them the Answer?
The true test of understanding is what a child can explain, teach, present, perform, or discuss live.
Nov 11, 2026 · 6 min read