Global AI Watch: Mapping the School, Seeing the System: How Spatial Context Reshaped Public Decision-Making in Uzbekistan and Bhutan
Read the interview between Aziza Umarova and Beth Simone Noveck
We did not set out to build another digital dashboard.
The question that guided us in Uzbekistan was practical: how do existing data systems actually see schools, and what isn’t visible once the schools are seen?
This work emerged under conditions familiar to many public institutions, a predominantly top-down governance context, fragmented administrative systems, limited resources, uneven data quality, and strong pressure to demonstrate results. These conditions ultimately shaped both the methods and the choices described in our approach.
In 2023, information about schools, clinics, roads, and public services was scattered across many places. Ministries kept their own datasets. Analysts produced careful reports. Indicators were assembled and elevated to the national level.
The data was mostly correct, yet what it described floated above the ground.
Distance, terrain, building condition, and geographic constraint rarely entered the picture. What mattered on paper did not always match what mattered in place.
That disconnect became visible the moment schools were mapped.
In both Uzbekistan and Bhutan, children leave home before sunrise, walking across uneven ground and along unmarked routes, navigating distance and terrain as part of their everyday access to education.
Once schools appeared on a map, they stopped behaving like rows in a table. They became points in landscapes, shaped by hills, rivers, roads, and settlement patterns.
Travel time replaced straight-line distance. Catchment areas emerged where none had been drawn. Panoramic images captured alongside spatial data revealed details such as a cracked wall, an unsafe entrance, and a long walk across uneven ground.
From Mapping to a Geoportal
Those early insights shaped the geoportal we later developed in Uzbekistan. The system was built on a simple idea that geographic context is not decoration, but structure.
By layering administrative records with spatial data and field-level observations, the geoportal allowed schools to be seen not only as existing or non-existent, but also as accessible or distant, maintained or fragile, clustered or isolated. Schools served as a starting point, a place to test how data could be integrated, validated, and interpreted when location was treated as essential rather than optional.
The first effects were methodological, not institutional. Spatial visualization changed the kinds of questions analysts could ask.
Questionnaires developed for spatial mapping enabled the capture of dimensions that shape how education is actually experienced but are often absent from aggregate reporting, such as whether children with disabilities can physically reach classrooms, whether water, sanitation, and hygiene facilities are functional, and whether classrooms are overcrowded.
Questionnaire design became the practical link between global goals and everyday school realities.
These dimensions align with the intent of Sustainable Development Goal 4 on inclusive and equitable quality education, with important measurement advances prioritized by national education strategies, shifting attention away from counting enrollment and toward access, quality, and dignity.
Following the initial focus on schools and kindergartens, the geoportal gradually extended beyond education to encompass health, water, and other sectors.
The Tech Stack and Collective Data Collection
From a technical perspective, our team deliberately chose to build the platform entirely on open-source technologies, including QGIS, and to develop its analytical capabilities in-house using local technical resources.
This enabled the integration of advanced machine-learning and artificial intelligence components for pattern detection, validation, and scenario-based analysis. We relied on capabilities still uncommon in many public-sector systems, including using image analysis and scoring, based on agreed parameters, to flag infrastructure risks and help determine what to prioritize for follow-up and investment.
The approach prioritized reproducibility, long-term sustainability, and institutional ownership, avoided vendor lock-in, and lowered barriers to learning and adaptation. The platform was designed not as a fixed tool but as a living system that can evolve with the needs and capacities of the institutions that use it.
Data collection itself became participatory. Between 2023 and 2025, nationwide mapathons brought hundreds of young people into building and validating the geoportal. In 2025, more than 300 students from leading universities helped create a digital twin of community-level social infrastructure, mapping schools, kindergartens, healthcare facilities, and water access across local neighborhoods (mahallas) using mobile phones and standardized surveys.
From Place-Based Insight to Policy and Practice
From the outset, the Agency for Strategic Development and Reforms under the President of Uzbekistan played a central role in advancing this approach, translating evolving strategic planning requirements into shared analytical frameworks. Spatial methods were applied not as a standalone technical solution, but as an enabling layer across policy domains, with information made accessible to citizens for transparency and validation.
By revealing service gaps at the facility level, the geoportal directly informed the design of the centrally funded Clean Hands programme launched in May 2025, which allocates dedicated funding for water, sanitation, and hygiene across schools, preschools, and healthcare facilities. The same spatial logic shaped a national public investment management framework, enabling investment decisions to be assessed geographically rather than sector-by-sector.
As these practices matured, the geoportal was embedded in the country’s strategic planning architecture — most notably Uzbekistan–2030— linking plans and monitoring frameworks to location intelligence and opening data to citizens and other stakeholders for validation. Throughout, the Agency prioritized interoperability, fostering a shared spatial language for connected, place-based decision-making across institutions.
The experience gained traveled, informing similar work undertaken in the Kingdom of Bhutan. At the initiative of the Royal Civil Service Commission of Bhutan, and in collaboration with the National Statistics Bureau of Bhutan, my colleagues and I supported technical capacity-building efforts, introducing school mapping and geoportal workflows.
As before, once schools appeared on the map, geography began to speak. In Bhutan, this experience is now informing growing interest in using geoportal-based workflows within the National Statistics Bureau for data collection and validation, including preparations for the 2027 population and housing census.
Across both contexts, the same takeaways endured.
Data did not acquire meaning by existing alone, but by finding its place, in the footsteps of children walking to school, in the routes families take across hills and roads, in the rhythms of everyday life.
When placed back into the world it describes, data becomes more intelligible, not louder, but clearer.