Today, Eedi is delighted to announce that we are joining the ShikshaNext EdTech Accelerator and beginning our first exploratory study in India.
Through the ShikshaNext EdTech Accelerator, we're offering a WhatsApp-based learning experience to around 1,000 students in grades 6 to 8 in Rajasthan government schools. The student receives a short diagnostic quiz on their parent's WhatsApp, with topics aligned to what the teacher is covering in class. When they get something wrong, our constrained AI tutor activates to help them resolve their specific misconception. Constrained means the AI Tutor only engages with the student on a wrong answer, and only on that misconception. There is no open-ended chat.
We're running this as an exploratory study. Every student receives a weekly multiple-choice homework quiz on WhatsApp; a randomly chosen half also gets the AI tutor that steps in when they answer a question wrong. That lets us compare the quiz plus tutoring against the quiz alone.
What we're trying to learn in Rajasthan is how misconceptions get resolved in a setting like this, and through WhatsApp. We want to hear from teachers directly, and through interviews we'll learn how this homework-based approach fits with their work in the classroom.
At Eedi, we only scale what works. This exploratory study builds on numerous Eedi RCTs focused on diagnosing and resolving misconceptions over the last 5 years that provided gold-standard evidence of efficacy in UK classrooms. Of course, we'd love to replicate these positive results in India, but context is everything. Adapting our constrained AI tutor to the language and cultural needs of teachers and students in Rajasthan means developing new misconception mappings for their curriculum, using WhatsApp as the delivery surface for the first time, ensuring our safety and moderation framework is validated for Hindi, and, over the course of this work, building voice-first interaction for students who can't yet read fluently.
We're ready to start, and to learn from every challenge and opportunity the study presents. If the evidence is weak, we'll publish it anyway and share what we’ve learned. If it's strong, we'll scale the study, carefully, and only to the extent that the evidence supports it.
Thank you to ShikshaNext and Central Square Foundation for partnering with us on this important project, and to BCG and Google DeepMind, who have been in this with us from the design stage.
