This report presents findings from an 8-week pilot of the Kalimani model, an AI-enabled sign language tool for numeracy instruction. It covers teacher training, coaching, learner engagement, and numeracy outcomes across 12 schools.
This document is a learning report on the Kalimani Inclusive Numeracy project, a pilot study in Kenya and Tanzania. It is intended for educators, policymakers, and researchers interested in inclusive education for deaf and hard-of-hearing learners. The report presents evidence on the feasibility and acceptability of an AI teacher-support model that uses sign language visual demonstrations to boost numeracy skills in Grade 2 and 3 learners.
The executive summary outlines the Kalimani model, which includes government support, teacher professional development, and an avatar teaching numeracy in Kenyan and Tanzanian Sign Language. Key findings show that teacher training improved inclusive pedagogy knowledge from 30% to 90% and digital confidence from 37% to 75%. Teacher compliance with inclusive practices rose significantly in intervention schools, while control schools showed minimal change.
Learner numeracy improved in both intervention and control groups, with the largest gains among those starting at the lowest baseline. However, applied problem-solving domains remained low. Learner engagement increased strongly in intervention schools, reaching 82% to 90% across grades. The report includes qualitative insights from teachers, noting increased participation and collaboration.
The introduction discusses the systemic barriers faced by deaf learners in sub-Saharan Africa, including limited teacher training in sign-supported instruction. It highlights the importance of foundational numeracy as a gateway skill and the reliance on oral teaching methods that exclude deaf learners. The Kalimani innovation is described as an AI-enabled signing tool that operates through a continuous cycle of teacher-led lessons, tablet-based assessments, caregiver reinforcement, and…
The methodology section details the mixed-methods, non-randomised feasibility design over an 8-week pilot. Data were collected via Kobo Toolbox across 12 purposely selected government schools, with 309 learners assessed at baseline and 319 at endline. The report includes tables and figures on sample distribution, numeracy skills, and qualitative insights from Kenya and Tanzania.
The report concludes that the pilot offers credible early signals of model feasibility, aligning with SDG 4 and the Continental Education Strategy for Africa. Coaching through ministry structures emerged as a critical lever for embedding inclusive pedagogy. The findings point to the need for a larger, longer study with matched groups to attribute learning gains to the model.