Interest-based learning: what the research actually shows
Frederik Bjørnbak, Founder & CEO, GradeAid: former teacher in Denmark's public schools
10-second summary
Interest-based learning, or context personalisation, keeps the learning goal fixed and wraps the task in something the kid already cares about. A 2024 meta-analysis found medium effects on interest and retention, while the evidence for transfer is thin. A Norwegian study found no gain for the least interested kids, and a February 2026 study found teachers got better at writing personalised tasks with AI but not faster. The post ends with five steps a school can try in one class this term.
66 percent. That is the share of Danish 15-year-olds who say they are curious about many different things, against 73 percent across the OECD, and the Danish share has fallen 8.7 percentage points since 2022 (PISA 2025 Denmark country note, OECD, September 2026). The same report put Danish maths at 471 points and reading at 460, both lower than in any earlier PISA round. Interest-based learning is the idea that you take the thing a kid already cares about and build the task around it. It is being sold hard right now, often by EdTech companies like ours. So this post tries to say plainly what the research shows it does, and what it does not.
The curiosity number matters because the OECD also measured what it is worth. In Denmark, one unit more on the curiosity index went with 30 more points in science, against 19 points across the OECD (same report, September 2026).
Motivation falls as kids get older, in Norway too
Norway asks its kids directly. In the 2024 national pupil survey, 59 percent in grade 7 and 55 percent in grade 10 said they were interested in learning at school in many or in all or most subjects. In the first year of upper secondary, where kids have chosen a direction, the share was 67 percent (Udir, 2025). Grade 7 has been falling since 2019 and grade 10 since 2020.
We talk about this as if it were a phone problem, a parenting problem, a generation problem. I taught in Denmark's public schools, and what I saw in the room was simpler. A kid who knows every squad in the Premier League but freezes at a worksheet on percentages. A kid who has read 400 pages on dinosaurs and will not read the two pages in the textbook. A kid who edits videos for hours and says she cannot concentrate. The capacity is there. The lesson does not touch it.
What interest-based learning actually is
Researchers call it context personalisation. The maths stays the same. The percentages task is now about a transfer fee. The reading passage is now about a dinosaur. The goal did not move. The wrapper did.
It is not the same as learning styles, and we will cover that one separately. Interest-based learning claims something narrower: a kid attends longer and tries harder when the content connects to something they already know and like. That narrower claim has evidence behind it.
What the research shows: a 2024 meta-analysis
The best single summary is a meta-analysis by Lijia Lin, Xin Lin, Xiaofang Zhang and Paul Ginns in Educational Psychology Review (Lin et al., August 2024). They pooled 34 publications. Across 26 effect sizes and 5,335 participants, personalising learning to interests had a medium to large effect on interest (g = 0.55). Across 46 effect sizes and 5,991 participants, it had a medium effect on retention, meaning how much was remembered (g = 0.48). It also lowered reported cognitive load (g = 0.54, from 8 effect sizes).
The effect on interest depended on grain size, meaning how precisely the task matched the particular kid. A task about sport in general does less than a task about the kid's own club.
Behaviour moved as well. Candace Walkington and Matthew Bernacki followed 150 grade 11 kids in an algebra course over four months, 99 with problems rewritten around interests from a survey and 51 with the standard problems (Walkington and Bernacki, International Journal of Artificial Intelligence in Education, 2019). The personalised group learned more and gamed the system less, meaning fewer kids guessed their way through or fished for hints.
Where interest-based learning falls short
Now the awkward part. Transfer, using what you learned on a new kind of problem, rested on only 6 effect sizes and 375 participants in the meta-analysis, with a smaller effect (g = 0.36). That is thin evidence for the thing we most want.
A Norwegian study is a useful cold shower. Sigve Høgheim and Rolf Reber at the University of Bergen tested context personalisation and example choice with lower secondary kids in maths (Høgheim and Reber, Contemporary Educational Psychology, 2015). Personalising the context did not raise situational interest. Letting kids choose the example had a small effect, and mainly for kids who were already interested or already felt competent. The kids the method is meant for gained the least.
And the method costs effort. In a study posted in February 2026, Walkington and colleagues watched middle school maths teachers write personalised problems with ChatGPT (Walkington et al., February 2026). Teachers personalised at a broad level, such as sport or animals, while kids wanted fine detail and specific popular culture. Teachers spent their time fixing references, depth and realism. They got better at it. They did not get faster.
So interest does not do the teaching. It gets the kid to the table and keeps them there a little longer. The explanation, the worked example, the practice and the feedback are still the teacher's job, and I know of no study that shows otherwise.
Interest-based learning changes whether a kid starts the task and how long they stay with it, not whether the task teaches anything.
Three things worth taking from the evidence:
- Interest reliably raises interest and recall, with medium effects across 34 publications (Lin et al., August 2024).
- The effect shrinks when the match is vague and can vanish for the least interested kids (Høgheim and Reber, 2015).
- Instruction still has to be there. Interest is the wrapper, not the lesson.
What a school can do
This is what we would try in one class this term. The time costs are honest estimates.
1. Ask. One sheet, 10 minutes in a lesson: three things you do outside school and one thing you know a lot about. Keep it in the class folder. The February 2026 study found kids want the specific thing, so write down the club, the game and the series, not the category.
2. Rewrap one weekly task, not the week. The word problems in maths, the reading passage in a language lesson, the data set in science. About 20 minutes once to set it up, then 10 minutes a week.
3. Keep the goal fixed and visible. Write the objective for the lesson at the top of the rewrapped task, exactly as it stands in the class plan. If the interest version does not practise the same skill, bin it.
4. Watch two kids, not the class. Choose the two who start last. Note whether they start sooner and stay longer over four weeks. That is better evidence than anything in this post, because it is about your room.
5. Do not expect transfer for free. After the interest-wrapped practice, give one plain task on the same skill. If they cannot do it plain, the wrapper did its job and the teaching still has to be done.
Where GradeAid fits
GradeAid is a Danish platform for teachers. The teacher types in the topic and the level and gets a lesson plan, a worksheet, a quiz, a presentation and screen-free activities in seconds, which you can also differentiate. The material can be printed. The teacher can add a kid's interest to the topic, which is step 2 above, but the result still has to be checked against the goal: 47.6 percent of our worksheets get edited after generation (our own numbers, 14 September 2026), and we think that is the teacher in the loop doing the job.
If you work with kids who have stopped learning with interest, we would like to hear from you. Write to us at founders@gradeaid.ai.
Frequently asked questions
What is interest-based learning?
Interest-based learning, which researchers call context personalisation, keeps the skill and the learning goal fixed and changes the setting of the task to something a kid already cares about. A percentages task becomes a transfer fee, a reading passage becomes a dinosaur. The goal in the class plan does not move. Only the wrapper around the task does.
Does interest-based learning improve results?
Partly. A 2024 meta-analysis of 34 publications in Educational Psychology Review found medium effects on interest (g = 0.55) and on retention (g = 0.48). Evidence for transfer to new kinds of problems rested on only 375 participants and was smaller (g = 0.36). It raises attention and recall. It does not replace explanation, worked examples and practice.
Is interest-based learning the same as learning styles?
No. Learning styles claim that a kid learns better through a preferred channel, such as visual or auditory. Interest-based learning claims something narrower: that a kid attends longer and tries harder when the content connects to something they already know and like. The second claim has peer-reviewed evidence behind it, including the 2024 meta-analysis by Lin and colleagues.
How can a teacher use interest-based learning with a whole class?
Collect interests once on a single sheet, about 10 minutes in a lesson, and write down the specific club, game or series rather than a category. Then rewrap one task a week, such as the word problems in maths, and keep the lesson objective printed at the top. Watch the two kids who usually start last and note whether they start sooner over four weeks.
Does interest-based learning work for every kid?
Not equally. A Norwegian study by Høgheim and Reber (2015) found that personalising the context did not raise situational interest in lower secondary maths, and that letting kids choose examples helped mainly those who were already interested or felt competent. A vague match does little. The more precisely the task fits the particular kid, the larger the effect on interest.