RobotsMALI shows how AI-generated mother-tongue literature could expand literacy, support inclusive education, and strengthen linguistic identity in low-resource communities
In early 2025, UNESCO’s Global Education Monitoring Report flagged mother-tongue instruction as one of the most underfunded yet highest-impact levers for reducing the global learning crisis, where over a quarter of a billion learners lack instruction in a language in which they are proficient. The institutionalisation of native languages as mediums of instruction is essential for combating illiteracy among their speakers; however, this process hinges critically upon the availability of a diverse literary corpus. A scarcity of literature in vernacular languages perpetuates educational barriers, as learners struggle with unfamiliar dominant languages, causing disorientation and high dropout rates. Such literature provides authentic exposure to linguistic variation, cultural contexts, and sociolinguistic nuances, thereby fostering reading proficiency, critical thinking, and cultural identity.
However, generating such corpora demands extensive time and resources, particularly for low-resource languages that are underrepresented in print. Artificial intelligence (AI) offers a transformative solution, enabling rapid, scalable production of culturally attuned literature through generative models and translation pipelines. AI tools like large language models can craft stories, translate them into target languages, and pair them with relevant imagery, all under human oversight to ensure accuracy and sensitivity.
After Mali shifted in 2023 from French to local languages such as Bambara, RobotsMali used AI to produce 107 culturally grounded children’s books in under a year, distributed to over 300 pupils, boosting literacy, easing comprehension, and preserving linguistic heritage.
RobotsMALI represents one of the best-documented examples of AI-powered mother-tongue education. After Mali shifted in 2023 from French to local languages such as Bambara, RobotsMali used AI to produce 107 culturally grounded children’s books in under a year, distributed to over 300 pupils, boosting literacy, easing comprehension, and preserving linguistic heritage. The timing of this analysis is significant, as falling AI inference costs, improving multilingual models, and growing donor interest have created a narrow window to evaluate, refine, and scale this model before fragmented, underfunded pilots entrench poor practices.
Mali grapples with an illiteracy rate of close to 64 percent (as of 2024), and a primary cause is the entrenched use of French as the primary medium of instruction despite its limited presence in households. This linguistic disconnect stems from Mali’s colonial history under French rule from the late nineteenth century until independence in 1960, which imposed French as the administrative and educational lingua franca while marginalising indigenous languages.
Compounding this challenge is Mali's dire socio-economic context: it ranks among the lowest on the Human Development Index, hampered by low levels of industrialisation and ongoing security threats from jihadist groups operating in remote borderlands. The nation recognises thirteen national languages, with Bambara spoken by approximately 80 percent of the population. However, these remain predominantly oral traditions, with literacy confined to a negligible fraction of speakers owing to the near absence of written literature. This scarcity perpetuates a cycle in which foundational mother-tongue education remains unimplementable, as viable reading materials are largely unavailable to incentivise written proficiency.
The AI for Education project by RobotsMALI proposed a systemic solution centred on rapidly generating culturally resonant literature to establish Malian languages as viable mediums of instruction, foster a rapidly developing literary identity, and encourage literacy acquisition.
To address this, the AI for Education project by RobotsMALI proposed a systemic solution centred on rapidly generating culturally resonant literature to establish Malian languages as viable mediums of instruction, foster a rapidly developing literary identity, and encourage literacy acquisition. Artificial intelligence proved instrumental in this process, accelerating content production through a chain of tools that translate author-drafted narratives from English or French into local languages such as Bambara. Critically, AI enabled hyper-personalised materials by modulating plot complexity, vocabulary, sentence length, grammatical structures, and cultural settings to suit diverse reading levels, which is expected to be especially valuable in teacher-scarce regions. Authors crafted stories optimised for translation and supplemented them with AI-generated pedagogical aids such as comprehension exercises, thereby constructing a scaffold for scalable, effective mother-tongue education.
Central to the global scalability of RobotsMALI’s AI for Education programme design is the ‘Windows and Mirrors’ approach, a pedagogical framework that aims to produce a comprehensive spectrum of reading materials spanning pre-school to high-school levels. As a result, the literature produced was also intended to cover a wide range of subjects to address and encourage varied interests among readers.
The ‘window’ dimension introduces learners to the broader world, broadening horizons beyond local confines and fostering global awareness. Conversely, the ‘mirror’ aspect anchors content in familiar cultural contexts, rendering it relatable and comprehensible, thereby enhancing engagement and retention. A practical example involves adapting renowned English or French narratives, such as classic fables, by substituting Eurocentric elements with Malian equivalents—for instance, replacing Alpine settings with Sahelian landscapes or European protagonists with Bambara-speaking villagers—before translating them into local languages.
This dual approach is critical for cultural consolidation and reinforcing linguistic identity, as it validates indigenous experiences while connecting them to universal themes. Moreover, leveraging an existing corpus of adaptable foreign stories streamlines material sourcing and accelerates the creation of high-quality literature in resource-constrained environments.
Deploying generative AI for literary production in low-resource languages introduces multifaceted challenges that demand meticulous mitigation strategies. Foremost among these is the imperative for rigorous human review and iterative revision to safeguard content quality, cultural appropriateness, and pedagogical efficacy. AI-generated narratives often require substantial editing to eliminate factual inaccuracies, tonal inconsistencies, or culturally insensitive elements. The scarcity of proficient writers fluent in these languages exacerbates this issue, rendering quality assurance both time-consuming and resource-intensive.
The overwhelming majority of the approximately 1,000 languages spoken across Africa lack substantial digital footprints and are severely underrepresented in the datasets on which large language models are trained.
Compounding these hurdles is the inherent bias in AI training data. The overwhelming majority of the approximately 1,000 languages spoken across Africa lack substantial digital footprints and are severely underrepresented in the datasets on which large language models are trained. Consequently, outputs for these languages frequently exhibit inaccuracies in grammar, idiom, and cultural nuance. Such systems may perform adequately in Global North contexts but falter elsewhere, necessitating sophisticated prompt engineering to produce viable results.
Furthermore, incomplete or emergent writing systems for many such languages hinder the development of natively trained generative models, confining advanced AI capabilities to speculative horizons. These models remain tethered to translation pipelines from high-resource languages, perpetuating dependency and limiting true linguistic autonomy. Despite these obstacles, hybrid human-AI workflows demonstrate viability, where targeted interventions transform nascent tools into potent instruments for linguistic revitalisation.
Across Sub-Saharan Africa, South Asia, and Southeast Asia, governments are increasingly reversing the colonial-era dominance of European languages in education. In 2023, Tanzania reinforced Swahili as its primary medium of instruction. Ethiopia has long mandated regional languages such as Afaan Oromoo and Amharic in primary schooling. In India, the National Education Policy of 2020 mandates mother-tongue instruction through at least Grade 5, creating urgent demand for vernacular content across dozens of scheduled and tribal languages. In Papua New Guinea, over 800 living languages coexist with near-total dependence on English in formal schooling, a structural mismatch that AI-assisted content production could begin to address.
Mali is distinctive for the speed and deliberateness of its AI deployment following a national policy shift, making it an unusually clean case through which to examine a replicable model.The AI for Education initiative establishes a robust precedent for leveraging AI-driven literary generation in tandem with script development, offering a replicable model for linguistic revitalisation worldwide. Its successful pilot that produced over 140 culturally attuned books in Bambara within a year demonstrates that hybrid human-AI workflows can rapidly construct literary corpora even in oral-dominant, low-resource contexts, thereby enabling mother-tongue instruction at scale.
Integrating script standardisation with AI-assisted literature production holds transformative potential, particularly for the Global South's marginalised linguistic communities.
This approach is adaptable to diverse educational contexts where linguistic gaps impede learning. A pertinent example is the Santhali language, spoken by over seven million people primarily in India, which adopted the Ol-Chiki script as an official writing system in 1925 — a comparatively recent development. Applying RobotsMALI’s methodology, AI could generate reading materials spanning elementary prose to advanced narratives in Santhali, with outputs refined to reflect local cultural motifs. This could accelerate the transition to Santhali-medium instruction, mirroring Mali’s literacy gains, and could be scaled to address India’s multilingual educational challenges.
Integrating script standardisation with AI-assisted literature production holds transformative potential, particularly for the Global South's marginalised linguistic communities. Across the rest of Africa, where hundreds of oral languages await codification, and among Indian tribal groups with nascent scripts such as Ol-Chiki, such synergies could help bridge oral-written divides, foster cultural identity, and combat illiteracy. By prioritising the ‘windows and mirrors’ approach alongside rigorous quality controls, these initiatives could promote equitable cognitive development and socioeconomic advancement, positioning RobotsMALI as a potential blueprint for inclusive, technology-enabled language policy worldwide.
The AI for Education initiative by RobotsMALI exemplifies how AI-driven literary generation can surmount longstanding barriers to institutionalising native languages as mediums of instruction, thereby addressing illiteracy among mother-tongue speakers. By rapidly producing culturally resonant corpora, ranging from pre-school primers to advanced narratives, this approach addresses the scarcity of vernacular literature that perpetuates educational exclusion, dropout, and cognitive disorientation.
As evidenced by the success of the pilot conducted by RobotsMALI, strategic AI deployment, coupled with script standardisation, can democratise education, preserve linguistic heritage, and cultivate the reasoning skills needed for the modern world.
The project’s scalability holds immense potential and could herald transformative prospects for the Global South by fostering literate, empowered populations capable of driving development. As evidenced by the success of the pilot conducted by RobotsMALI, strategic AI deployment, coupled with script standardisation, can democratise education, preserve linguistic heritage, and cultivate the reasoning skills needed for the modern world. Policymakers worldwide must seize upon this blueprint to bridge linguistic gaps and ensure equitable literacy as a cornerstone of inclusive progress.
Pranoy Jainendran is a Research Assistant with the Centre for Security, Strategy and Technology at the Observer Research Foundation.
The views expressed above belong to the author(s). ORF research and analyses now available on Telegram! Click here to access our curated content — blogs, longforms and interviews.
Pranoy Jainendran is a Research Assistant with ORF’s Centre for Security, Strategy & Technology. His work examines how technology shapes State institutions, national and international affairs, ...
Read More +
PREV
