Tutor children in maths over WhatsApp on a basic phone
Rising Academies' Rori tutors children in maths over WhatsApp on basic phones. In an eight-month study in 11 Ghanaian schools, pupils who used it twice a week made larger maths gains, with an effect size of 0.37.
Built withClaude
Try it withClaudeChatGPT / OpenAI modelsGeminiGrok
Deployed; used in Ghana, Sierra Leone, Nigeria, Kenya and Rwanda, with more than 150,000 students reached according to Anthropic's case study, Rising Academies (Rising Academy Network)
Why it matters
Rising Academies, an education network founded in Sierra Leone in 2014, says that education tools needing laptops and fast internet reach fewer than 5% of students where it works, according to Anthropic's case study. Rori runs on WhatsApp, on basic mobile phones and low-bandwidth networks. In the first-year study in Ghana, pupils who used Rori for two 30-minute sessions a week, on top of their normal lessons, made larger maths gains than pupils who did not.
How it works
Pupils message Rori on WhatsApp. It mixes pre-written lesson plans with conversations the AI generates, adjusts the difficulty of problems, and gives encouraging explanations that link maths to daily life and career goals. The design follows Teaching at the Right Level, an approach that groups children by the level they have reached. According to Anthropic's case study, Claude marks open-ended answers, recognising correct answers written in different formats, and uses retrieval to keep off-topic chats educational. In the Ghana study, about 1,000 pupils in grades 3 to 9 across 11 schools took part for eight months: pupils in the treatment group used Rori in two 30-minute sessions a week, and pupils in the control group carried on with their regular maths lessons. In classrooms, pupils worked problems out with pen and paper and asked teachers for help, which led the team to plan better hints for times when teachers are stretched. Rising Academies has also built Tari, a WhatsApp tool for teachers that starts conversations about lesson planning and checks in when a teacher does not reply.
Who it's for
Ministries of education, school networks and NGOs working where pupils have basic phones but no laptops, and parents or teachers who want structured maths practice for a child.
The prompt
You are a patient maths tutor for a child. Teach over chat, one short message at a time, as if on a phone. Do not ask for the child's full name, school or address. Child's age or school year: [for example, 10 years old, year 5] Topic to practise: [for example, fractions, place value, times tables] Language for explanations: [for example, English] 1. Start with three quick questions to find the level the child can do comfortably, from simple to harder. 2. Pick the level just above that and give one question at a time. Wait for the answer before moving on. 3. If an answer is right, say so briefly and make the next question a little harder. Accept correct answers written in different ways, such as 0.5, 1/2 or one half. 4. If an answer is wrong, do not give the answer straight away. Give one hint, then a second hint, and only then show the working step by step. 5. Every few questions, connect the maths to something from daily life, such as shopping, cooking or sharing. 6. After ten questions, finish with a short summary for the adult: what the child can do, what to practise next, and one question to try tomorrow. Keep the tone warm and encouraging. If the child asks about something unrelated or unsafe, gently bring the chat back to maths and suggest they talk to a trusted adult.
Open inClaude (opens in a new tab)ChatGPT (opens in a new tab)Gemini (opens in a new tab)Grok (opens in a new tab)
Sources
The Ghana study was written by researchers who include members of the Rori team and was posted as a working paper. Anthropic's case study is from Rising Academies' technology vendor. The J-PAL listing describes a separate pilot that has not reported results, and the Learning Agency interview adds detail on classroom use from the Rori team.
1. arXiv (Henkel, Horne-Robinson, Kozhakhmetova and Lee): "an AI powered conversational math tutor accessible via WhatsApp"; 11 schools, grades 3 to 9, about 1,000 students, eight months, two 30-minute sessions a week; effect size of 0.37; "only report on year 1 of the intervention". Effective and Scalable Math Support: Evidence on the Impact of an AI-Tutor on Math Achievement in Ghana (opens in a new tab)
2. Anthropic: "Already reached over 150,000 students"; "Ghana, Sierra Leone, Nigeria, Kenya and Rwanda"; "0.3 standard deviation effect size in learning outcomes"; "reach less than 5% of students in their service regions"; Claude evaluates open-ended responses and uses retrieval; Tari checks in when users don't respond. Rising Academies case study (opens in a new tab)
3. J-PAL: "We plan to recruit 24 schools that will be split into two conditions"; researcher Edward Asiedu; Rori accessed through WhatsApp; no results reported. AI, Implementation and Academic Outcomes in Low-Resource Settings: Experimental Evidence from Ghana (opens in a new tab)
4. The Learning Agency: Rori runs on WhatsApp for low-bandwidth settings; adjusts problem difficulty; modelled on Teaching at the Right Level; pupils used pen and paper and asked teachers for help. 5 Questions with the Developers of the Math Chatbot Rori (opens in a new tab)