A new study published in Frontiers in Sustainable Food Systems reveals that gender disparities in the use of smartphone-based artificial intelligence tools for banana disease management in Rwanda are driven primarily by unequal access to devices and infrastructure, rather than by differences in technical capability. The research, which utilized the Tumaini app as a case study, investigated how gender shapes access to these digital advisory tools among smallholder farmers. By distinguishing between the structural conditions that govern access and the technical competencies required for use, the study challenges the assumption that low adoption rates among women reflect a deficit in skill or willingness to learn.
The investigation employed a mixed-methods design to capture both quantitative and qualitative dimensions of digital adoption. The research team conducted a survey of 357 smallholder farmers across three districts in Rwanda: Gasabo, Karongi, and Rusizi. These districts were selected purposively because they have been significantly affected by Banana Bunchy Top Disease (BBTD), a condition that can cause yield losses of up to 100% and has no cure once a plant is infected. The study also included 12 focus group discussions with 72 participants, in-depth interviews with 61 farmers, and a 12-month longitudinal monitoring of WhatsApp engagement among 40 smartphone owners.
The results indicated that men had significantly greater smartphone ownership than women, with 24% of men owning smartphones compared to only 6.5% of women. This disparity persisted even after controlling for education, income, and electricity access. A binary logistic regression analysis showed that male farmers had 4.47 times higher odds of owning a smartphone than female farmers, suggesting the presence of structural gender-related constraints. In contrast, among those who did own smartphones, no statistically significant gender differences were observed in digital literacy, indicating that women who had access performed core and advanced smartphone functions as well as or better than men.
Decision-making power within households also emerged as a significant factor. The study measured this using the Decision-Making Power Index (DMPI), which assesses autonomy in selling bananas, purchasing inputs, and spending income. Married women recorded a significantly lower mean DMPI of 46.95% compared to 64.63% for married men. This gap was particularly pronounced in the urban district of Gasabo, where men’s decision-making power was significantly higher than in rural districts, while women’s scores remained relatively stable across locations. This suggests that urbanization and market access may widen intra-household disparities if not accompanied by gender-responsive measures.
Language barriers further complicated access to the Tumaini app, which was available only in English, Kiswahili, and French at the time of the study. Kinyarwanda, the primary language of all participants, was not supported. Literacy in the app-supported languages was extremely low, with only 6.8% of women able to read or write in any of the three languages, compared to 12.2% of men. This mismatch between the tool’s interface and the users’ linguistic capabilities created a fundamental accessibility barrier, particularly for single women who had the lowest literacy rates in both Kinyarwanda and the app-supported languages.
Despite high stated willingness to adopt smartphone-based disease detection tools, with 85.4% of respondents expressing interest, actual post-training use remained limited. Among the 61 farmers who received training on Tumaini, only 39.3% used the app after the program. A significant gender gap was observed in this adoption, with only 26.3% of women using the app compared to 60.9% of men. The most frequently reported barriers for women included the unavailability of smartphones at the time of training, inability to navigate the app independently, and language difficulties in interpreting results. For men, the primary barriers were internet costs and installation challenges.
The 12-month WhatsApp monitoring provided objective data on actual usage behavior. Men were active on a mean of 56.0 days per year, compared to 34.8 days for women, a statistically significant difference. Men also sent more photographs and uploaded more diagnosis results. However, when engagement was normalized by the number of active days, the intensity of use was comparable between the sexes. Women sent approximately 0.36 photographs per active day, similar to the 0.32 sent by men. This suggests that when women did engage with the tool, they did so with similar intensity, but their overall participation was constrained by fewer opportunities and access to the device.
The study highlights that the gap between stated willingness and actual adoption reflects persistent structural barriers rather than a lack of motivation. Even when awareness and motivation are present, the lack of device ownership and limited control over resources hinder effective adoption. The findings indicate that interventions focused solely on training are unlikely to achieve equitable uptake unless they also address smartphone ownership, intra-household decision-making power, and navigation barriers embedded in tool design. The high willingness to adopt observed across both sexes indicates substantial latent demand that could be realized through targeted structural interventions.
The researchers recommend gender-responsive interventions that prioritize equitable access to devices, localized design in local languages, and structured group-based support. Providing smartphones through women’s groups and offering voice-guided interfaces in Kinyarwanda could help bridge the digital divide. The study emphasizes that closing the access gap is a food-security intervention as much as a gender-equity one, as every farmer excluded from timely diagnosis represents a potential locus of undetected disease spread and avoidable yield loss.
Limitations of the study include the small number of women who owned smartphones, which means that estimates for women’s digital literacy rest on a limited sample. Additionally, digital literacy was self-reported, which may understate women’s actual skills. The study was conducted in three specific districts in Rwanda, and while the findings are consistent with broader sub-Saharan African evidence, they may not be directly generalizable to other regions without further testing.
Despite these limitations, the study provides a detailed analysis of how gender intersects with socio-economic, cultural, and technical factors to shape access to AI-powered crop protection tools. It offers a framework for understanding the distinction between access and ability, which is transferable to other digital agricultural platforms. The findings underscore the need for inclusive digital transformation in agriculture, ensuring that women farmers can benefit equally from improved banana disease management and sustainable production.
The research contributes to the growing body of evidence on the digital divide in agriculture, highlighting the importance of addressing structural barriers alongside digital skills. By identifying the specific factors that constrain women’s access to AI tools, the study provides actionable insights for policymakers, developers, and development organizations. It suggests that a multi-sectoral approach, involving researchers, technology developers, and local communities, is essential for creating effective and equitable digital interventions.
As AI-enabled disease diagnosis becomes increasingly integrated into agricultural support systems, ensuring equitable access should be a central objective of digital agricultural policy. The study’s findings demonstrate that technical performance alone is insufficient for achieving equitable outcomes; user-side conditions, including device ownership, internet connectivity, and language accessibility, are critical determinants of adoption. Addressing these interconnected barriers is essential to ensure that AI-powered tools contribute to equitable agricultural transformation rather than reinforcing existing inequalities.
The study’s sampling strategy targeted smallholder farmers growing banana as a main crop, with the individual farmer as the primary unit of analysis. Researchers employed a multi-stage procedure, selecting sectors and cells in consultation with local agricultural officers, followed by the purposive selection of 12 villages based on the presence of BBTD and farm size. Farmers were then selected via systematic random sampling from lists provided by farmer promoters, stratified by gender and marital status. The final survey sample comprised 357 participants, adjusted from a target of 384 due to farmer unavailability during data collection. This sample included married men, their married counterparts, and women heading single-family households, ensuring a representation of different household structures and decision-making dynamics.
Internet access among the surveyed farmers was low overall, standing at 10.1%, and was shaped primarily by gender rather than by district or marital status. One in five men reported internet access, compared with fewer than one in twenty women, a highly significant gap confirmed by logistic regression. Among the 36 respondents with any access, use was most commonly weekly, followed by monthly and daily. This pattern indicates that the gender divide in digital connectivity is not merely a function of geographic location or household composition, but reflects deeper structural constraints on women’s access to digital infrastructure.
Power access at home was another critical factor, with about 60% of farmers owning an electricity supply and a further 6% using their own solar panel. Two-thirds of farmers had their own power source, while the remaining third relied on neighbors or had no source at all. Among those who relied on neighbors for charging, about one-third paid a small weekly fee, typically 100 Rwandan Francs. Married households were somewhat more likely to have their own power source than single-woman-headed households, though this difference did not reach statistical significance. Access to electricity did not vary significantly across districts, suggesting that power availability is a consistent challenge across the study sites.
The Tumaini app, which uses convolutional neural networks to diagnose banana diseases from photographs, was available in English, Kiswahili, and French at the time of the study. Kinyarwanda, the language of all participants, was not supported, creating a significant barrier for users with low literacy in the app’s languages. The app’s technical performance depends on user-side conditions, such as smartphone camera quality, which has been reported to reduce diagnostic accuracy by up to 40% for low-quality images. Improper image framing also accounted for about 30% of field misdiagnoses in one evaluation, highlighting the importance of user training and device quality in ensuring accurate diagnoses.
The study’s findings are consistent with broader evidence from sub-Saharan Africa, where infrastructure, affordability, and gendered access, rather than farmer capability, are identified as the binding constraints on digital-tool uptake. The “access, not ability” dynamic observed in Rwanda is unlikely to be country-specific and merits testing across other sub-Saharan banana-producing systems. Interventions focusing solely on training without addressing structural access barriers will have limited impact, underscoring the need for a multi-sectoral approach that includes device provision, localized design, and community-level engagement with gender norms.
Subject of Research: Access, not ability: gender disparities in access to smartphone-based AI advisory tools for banana disease management among smallholder farmers in Rwanda
Article Title: Access, not ability: gender disparities in access to smartphone-based AI advisory tools for banana disease management among smallholder farmers in Rwanda
Article References: Ndayizeye, J. P., Omondi, B. A., Kawarazuka, N., Bigirimana, V. D. P., & Gaidashova, S. (2026). Access, not ability: gender disparities in access to smartphone-based AI advisory tools for banana disease management among smallholder farmers in Rwanda. Frontiers in Sustainable Food Systems, 10, Article 1990189. https://doi.org/10.3389/fsufs.2026.1990189
Image Credits: AI Generated
DOI: 10.3389/fsufs.2026.1990189
Keywords: Access, ability, gender, disparities, smartphone-based, advisory, tools, banana, disease, management, smallholder, farmers
News Source: Alan Morgan. (November 11, 2026). Rwanda Study Links AI Tool Gender Gap to Access, Not Skill. Scienmag.
