The encyclopedia · Product & Design · Product decision · 2015–2019
Adaptive Mindspark software lifted Delhi students' math scores 0.37 standard deviations
An after-school software program that meets each child at their level outperformed one-size-fits-all classroom teaching in a randomized trial.
Educational Initiatives
the move
Educational Initiatives built Mindspark as an adaptive learning platform for Indian schools. J-PAL researchers Muralidharan, Singh and Ganimian ran a randomized evaluation with 619 secondary school students in Delhi, giving lottery winners access to 90-minute after-school sessions on the software.
Students offered the voucher scored 0.37 standard deviations higher in math — improving by more than twice as much as the comparison group — and 0.23 standard deviations higher in Hindi. The program raised test scores across all groups and was cost-effective compared with traditional schooling models.
The mechanism is adaptation: the software tracks each child's errors and misconceptions and adjusts difficulty and content, so instruction matches the student's level rather than the syllabus. The study was published in the American Economic Review in 2019.
why it works
- Content adapts to each child's actual level.
- Extra practice time compounds the adaptive benefit.
- RCT evidence convinced schools and funders.
- Software scales to millions without new teachers.
what transfers
Classrooms fail because one lesson level can't fit forty learners; adaptive software gives each child a personal path. Prove it with an RCT — measured gains made the model impossible to wave away.
what came after
The trial became a reference point for education technology, and EI continued to deploy Mindspark across Indian schools and state programs while extending the platform into foundational-literacy work for younger learners.
references
- Disrupting Education? Evidence on Technology-Aided Instruction in India (evaluation)
- Disrupting Education? Experimental Evidence on Technology-Aided Instruction in India
spotted an error? The archive wants to know.