Researchers asked a straightforward question: what happens when teenagers learn maths by analyzing mistakes in AI-generated solutions? The study tracked students as they worked through this approach, measuring both their maths performance and their ability to spot when AI gets things wrong.
The students were not identified by specific age or exact sample size in the available details, and the research comes from Frontiers in Education.
Math scores and AI error-detection both more than doubled
The results were striking. Students' maths scores jumped from an average of 2.78 to 6.39 out of 10—nearly a doubling. At the same time, their ability to detect AI errors rose from 3.61 to 7.72, more than doubling as well. The study notes these improvements had a large effect size, meaning the gains were substantial, not the result of chance.
Beyond test scores, students developed lasting independent habits. They began checking their own work regularly, recognized patterns in where AI commonly stumbles (fractions, decimals, percentages), and felt confident applying these verification skills in other contexts.
Error analysis and AI literacy reinforce each other
The key finding is that improvements in maths ability and AI error-detection happened together. Learning to critique AI wasn't a distraction from maths learning—it deepened it. The study suggests these two competencies strengthen each other rather than competing for attention.
For Singapore families, this matters because many teenagers use AI tools like ChatGPT for homework without any formal framework for questioning accuracy. This research points to a practical middle ground: rather than banning AI or treating it as a shortcut, families can teach teens to use it as a learning tool by actively critiquing it.
Reframing what "getting it right" means in the AI age
In education cultures that emphasize mastery and correct answers, this research introduces an important reframe. The real skill now is understanding *why* something might be wrong—whether it comes from a textbook, a classmate, or an algorithm. This builds the discernment teens will need to work alongside AI throughout their lives.
How to build this skill at home
- When your teen uses AI for homework, ask them to find three potential errors before accepting the answer.
- Praise them for catching mistakes in AI work, not just for getting the final answer correct.
- Help them notice patterns—which topics does AI struggle with most?—so they know when to double-check.