Conception d’un assistant numérique pour l’inclusion numérique rurale

Thesis Abstract

This study investigates how a digitally assisted interface can foster rural inclusion by bridging the persistent digital divide affecting underserved communities in agricultural regions. Despite increasing national connectivity, rural populations encounter barriers such as limited digital literacy, unreliable infrastructure, and socio-cultural factors that hamper effective ICT adoption. The aim is to design, implement, and evaluate a context-aware digital assistant tailored to rural users, with objectives to (i) identify user requirements across diverse rural demographics, (ii) develop an accessible, multilingual conversational assistant integrated with offline-capable features, (iii) assess usability, trust, and perceived usefulness, and (iv) evaluate impacts on digital participation, service access, and socio-economic indicators over a 12-month period. The study adopts a mixed-methods design underpinned by the Technology Acceptance Model (TAM) and the Diffusion of Innovations (DOI) framework. The population comprises rural adults in three distinct agrarian districts, with a target sample size of 420 participants for quantitative surveys and 60 in-depth qualitative interviews. A stratified random sampling approach ensures representation by age, gender, education, and digital literacy. Data collection instruments include a structured questionnaire measuring perceived usefulness, perceived ease of use, social influence, and behavioral intention, as well as semi-structured interview guides and task-based usability tests. The digital assistant prototype, named RurALink, integrates natural language processing for local languages, offline data caching, voice-enabled navigation, and a context-aware help system linked to essential services (health, agriculture, government portals). Validity and reliability are ensured through content validity panels with ICT4D experts and a pilot with 40 participants, yielding a Cronbach’s alpha above 0.85 for key scales. Quantitative data will be analyzed using multiple regression and structural equation modeling to test TAM and DOI relationships, while qualitative data will undergo thematic analysis guided by Braun and Clarke to extract patterns related to user experience, trust, and cultural fit. An integrative framework will synthesize findings across data strands, and a usage analytics module will capture interaction patterns, feature adoption rates, and offline vs. online activity. Expected findings include (i) high acceptance of RurALink among individuals with basic digital literacy when designed for local languages and offline capacity; (ii) significant mediation effects of perceived usefulness and social influence on intention to use; (iii) improved access to agricultural extension services, digital government portals, and health information; (iv) reduction in time previously spent on information-seeking tasks and a measurable increase in digital confidence. The study contributes to knowledge by operationalizing an ICT-driven inclusion solution tailored to rural contexts, expanding empirical evidence on digital assistant adoption in low-resource settings, and offering a scalable model for participatory design that integrates local languages and offline capabilities. Practical implications include a replicable development framework for rural digital assistants, guidelines for stakeholder engagement (goverment agencies, extension services, and community organizations), and policy recommendations to support infrastructure investments and digital literacy programs. The main conclusion anticipates that a context-aware, linguistically inclusive, offline-capable digital assistant substantially enhances digital participation and access to essential services in rural areas, provided that ongoing support, local governance involvement, and continuous content localization are maintained. Recommendations emphasize iterative co-design with rural communities, capacity-building initiatives for local facilitators, investment in resilient connectivity and caching strategies, and rigorous post-deployment monitoring to adapt features to evolving community needs.