On this page
Most Indian classrooms teach one syllabus to a room of children who are at very different levels. A study from Delhi asked a simple question: what happens if software teaches each child at the level they are actually at? This post walks through what the researchers did, what they found, and what it does and does not tell students, parents and teachers.
What question did the Mindspark study ask?
Karthik Muralidharan, Abhijeet Singh and Alejandro Ganimian wanted to know whether a "personalized technology-aided after-school instruction program" could raise learning in a short time. The paper, Disrupting Education? Experimental Evidence on Technology-Aided Instruction in India, was published in the American Economic Review in 2019.
The program was Mindspark, software built by the Indian firm Educational Initiatives. Its key feature is that it keeps testing a child and adjusts the next question to that child's level, including going back several grades if needed.
Who was studied, and how?
According to the working-paper version (NBER, revised July 2017), the sample was 619 students recruited from public middle schools in Delhi, about 97.5% of them in grades 6 to 9. Students applied for a place at Mindspark centres, and a lottery decided who got a free voucher: 314 were assigned to the treatment group and 305 to the control group. Control students were promised access after the study.
The centres ran six days a week for 90 minutes a day: 45 minutes on the software and 45 minutes of group support from a teaching assistant in batches of 12 to 15 students. The usual fee was Rs 200 a month, which the voucher covered. Learning was measured with independent paper-and-pencil tests in maths and Hindi before and after the 4.5-month programme.
A lottery matters because it makes the two groups alike on average, in motivation, family background and prior marks. Any difference at the end can then be put down to the programme rather than to who chose to join.
What did the study find?
The published abstract gives the headline numbers. Lottery winners scored 0.37 standard deviations higher in maths and 0.23 higher in Hindi than lottery losers after about 4.5 months. Using the lottery to estimate the effect of actually attending, the authors estimate that 90 days of attendance would raise maths and Hindi scores by about 0.6 and 0.39 standard deviations.
Three further findings from the working paper are worth knowing:
Children were far behind. The average grade 6 student was about 2.5 grade levels behind in maths, and by grade 9 about 4.5 grade levels behind. Within one grade, students spanned five to six grade levels.
Weaker students gained most in relative terms. The absolute gain was similar for everyone, but in the control group the bottom third of students made almost no measurable progress over the year. For them the programme was the difference between learning and not learning.
The gains were below grade level in maths. Because most children were so far behind, the software mostly gave them below-grade content. Maths gains showed up on below-grade questions, and there was no significant gain on grade-level school exams in maths, though there was in Hindi.
Attendance was not perfect: lottery winners attended 58% of the time on average, around 50 of 86 possible days.
What does the study not show?
Read the limits carefully before drawing big conclusions.
It is one city and a short period. The study ran for 4.5 months in Delhi with students who chose to apply. The authors discuss how far the results might extend to other students.
It was not software alone. Each session included 45 minutes with a teaching assistant, who mainly helped with homework and exam preparation. The study tests the whole centre model.
It does not say any app works. Mindspark had years of development and a large question bank tuned to Indian students' errors. A generic app without that diagnostic depth is a different product.
School marks did not move in maths. If you judge by the school exam, you might miss real learning that happened at a lower level.
What does it mean for Indian students and parents?
The main lesson is about level, not about screens. If a Class 9 student struggles with algebra because fractions never made sense, more Class 9 algebra will not fix it. The study suggests going back to the gap, even if it feels like "Class 6 work", and then building up. The related idea of mastery before moving on is explained in our post on mastery learning and adaptive practice.
For parents, a flat school report card can hide progress on basics. Ask what your child can now do that they could not do three months ago.
A second Indian study shows the other side: in Gujarat, computer practice that replaced class time made students learn less. How technology is used matters as much as whether it is used.
How to read a study like this
When you see "0.37 standard deviations", ask three questions. Compared with whom (here, lottery losers)? Over what time (4.5 months)? Measured on what test (an independent test that included below-grade questions)? The same number can mean very different things depending on the answers.
If you are working on a gap in a concept you were supposed to learn years ago, the Fermi Concept Library covers every concept in physics, chemistry, biology, mathematics and computing, so you can go back to the idea that is missing rather than rereading this year's chapter.
Pick one topic you find hard this week and find the earliest idea in it you cannot explain; start there.
