An AI maths tutor that adjusted to each student raised interest and test scores

The study

1. 01

   Who took part

   602 students in grades 7 to 10, about 14 years old on average, in 24 classes at public middle and high schools. About half were girls. 302 used the AI tutor and 300 kept their usual practice.

2. 02

   What they did

   The AI group did maths on a computer tutor 2 to 3 times a week for 30 to 40 minutes. This went on for about 10 weeks, or about 10 hours in all. The tutor chose problems to suit each student, gave feedback on each step and offered hints. It moved students on once they knew a skill well. The other group used books, worksheets or a plain problem website.

3. 03

   What happened

   At the end, the AI group rated themselves as more interested in maths than the other group. They also felt more sure of their skills. They gained a little less in effort and in how much they valued maths. They scored about 5 points higher out of 100 on a maths test. Parts of the paper's text are garbled, and it is unclear how students were split into groups in some classes.

What the study found

1. HelpedMotivation

   Maths on an AI tutor that adjusted to each student left students more interested and sure than usual practice.

   • Action: Use an adaptive study tooltested
   • Rule: Get feedback on your progresstested

   Randomised trial of 602 school students in grades 7 to 10 · vs usual textbook or worksheet practice · they rated themselves · ⚠ numbers unreliable

2. HelpedGoals

   After about ten weeks on the AI tutor, students also scored higher on a maths test than with usual practice.

   • Action: Use an adaptive study tooltested
   • Rule: Get feedback on your progresstested

   Randomised trial of 602 school students in grades 7 to 10 · vs usual textbook or worksheet practice · they rated themselves · ⚠ numbers unreliable

Source