At the forefront of the global push for sustainable development is Dominic Jainy, an IT professional whose work bridges the complex worlds of artificial intelligence and public policy. He specializes in transforming the vast, chaotic streams of modern data into actionable insights that can help governments tackle our most pressing challenges. In our conversation, we explore how this technological revolution is reshaping everything from climate action and gender equality to economic planning. We delve into the practicalities of turning satellite images and mobility data into tangible policy tools, the critical importance of building public trust through transparent governance, and the collaborative frameworks needed to ensure these powerful technologies accelerate, rather than merely measure, progress toward the Sustainable Development Goals.
The modern data ecosystem includes everything from satellite images to social media trails. How does AI specifically help convert these diverse inputs into usable signals for policymakers, and could you walk us through a step-by-step example of this process in action?
AI is essentially the critical linchpin that makes sense of this data deluge. Much of the information we get from satellites, sensors, or online platforms is unstructured; it’s messy, it arrives in torrents, and it’s not in a neat spreadsheet. AI’s power lies in its ability to perform pattern recognition and anomaly detection at an immense scale. It sifts through this noise to find the signal. For example, computer vision can analyze thousands of satellite images to quantify deforestation far faster than any human team, and natural language processing can scan public discourse to catch early warnings of social risks. This partnership allows human experts to stop drowning in raw data and start focusing on higher-level tasks like framing the right questions, validating the machine’s findings, and designing effective interventions.
Let’s take a disaster response scenario. First, AI models begin ingesting multiple data streams in near real time: high-resolution satellite imagery, social media posts from the affected area, and sensor feeds. Second, different AI techniques get to work—computer vision models parse the satellite photos to map flooded areas and damaged infrastructure, while natural language processing filters social media for urgent requests for help or reports of emerging health issues. Third, these processed layers of information are integrated and visualized on a single interactive map. A policymaker no longer sees just disconnected data points; they see a holistic, operational picture that allows them to move from hindsight to foresight, deploying resources precisely where they are needed most.
Tools that map sand and dust storm exposure by integrating hazard intensity with population data are incredibly powerful. How do such AI-driven visualizations change the way authorities plan for public health impacts and protect agriculture, and what are the key data streams required?
These visualizations are game-changers because they transform an abstract threat into a tangible, geographically specific risk. Before, an agency might issue a general warning for an entire region. Now, with tools like those developed by ESCAP’s Asian and Pacific Centre for the Development of Disaster Information Management, they can see exactly which communities, health clinics, or agricultural zones are in the direct path of the most intense part of a storm. This changes planning from a reactive exercise to a proactive one. Authorities can issue targeted health advisories to specific vulnerable populations, pre-position respiratory aid at clinics in the highest-risk zones, and advise farmers in precise locations on measures to protect their crops and livestock. It’s about precision and foresight.
To build these powerful risk layers, you need three key data streams. The first is the hazard data itself, which includes satellite and sensor readings that measure the intensity and trajectory of the sand and dust storm. The second is population data, which provides a detailed density map of where people live. Finally, you need land use data, which identifies critical areas like agricultural land coverage. By overlaying these streams, AI doesn’t just show where the storm is; it shows who and what is in its path, enabling a far more intelligent and effective response.
Combining household survey data with satellite-derived environmental measures like drought indices is an innovative approach. How does this linkage help reveal the hidden drivers of social outcomes like child marriage, and what challenges arise when integrating such disparate datasets for policy decisions?
This approach allows us to see how environmental context shapes social realities. For years, we’ve studied issues like child marriage through the lens of poverty, education, and social norms, which are all crucial. But by integrating household survey microdata with satellite measures like a vegetation index, we can test complex new hypotheses. For instance, UN Women’s work explores whether a prolonged drought in a specific district, visible from space, correlates with a local spike in child marriage reported in surveys. This can reveal that environmental stress acts as a powerful, often invisible, catalyst that interacts with and amplifies existing social vulnerabilities. It helps policymakers understand that progress on gender equality is fundamentally intertwined with climate resilience, leading to more holistic, multi-sector interventions.
The challenges are significant, however. First, you are merging completely different types of datfinely detailed, personal survey responses and broad, pixel-based satellite imagery. Aligning them spatially and temporally requires sophisticated data engineering. Second, there’s a major privacy consideration. You must link the datasets in a way that reveals geographic trends without ever compromising the anonymity of the household survey participants. Finally, establishing causality is difficult; the model can show a strong correlation, but it takes careful, domain-expert-led analysis to interpret what that correlation means and to design policies that address the true root causes rather than just the symptoms.
Using anonymized mobility data to generate real-time economic indicators is a major shift from traditional statistics. What governance frameworks are essential to ensure privacy while tracking tourism or retail footfall, and how can statistical agencies validate these new high-frequency proxies against official sources?
This is about making the economy “instrumented” so we can read it in real time, which is a massive leap from waiting months for traditional indicators. Indonesia’s work with tourism statistics is a fantastic example; by analyzing anonymized mobility patterns, they can approximate visitor flows and length of stay almost instantly. However, public trust is the keystone, and that trust hinges entirely on robust governance. The essential framework must include uncompromising protocols for anonymization to ensure no individual can ever be re-identified. It also requires clear standards for consent and transparent data-sharing agreements. The goal is to create what we call privacy-preserving signals, where the utility of the aggregate data is retained but individual privacy is guaranteed.
For validation, the key method is triangulation. A statistical agency can’t simply replace its old methods overnight. Instead, it must run the new and old systems in parallel. They would take the high-frequency mobility data for tourism and compare it against official, trusted sources like hotel occupancy records or flight arrival data over a historical period. By doing this, they can understand the new proxy’s strengths and weaknesses, quantify its margin of error, and build institutional confidence. This process ensures they can publish timelier estimates without sacrificing the rigor and credibility that is their core mandate.
Public trust is critical for the success of these initiatives. Beyond privacy protocols, what practical steps can an agency take to address potential bias in AI models and ensure accountability in their outputs? Please describe the key elements of an effective transparency framework.
You’re right, trust goes far beyond just privacy. It’s about believing the system is fair and accountable. One of the most important practical steps is to insist on model documentation. Every AI model used for a policy decision should come with what you could think of as a “nutrition label,” clearly stating what data it was trained on, its known limitations, and the assumptions baked into its algorithm. This prevents the model from being an impenetrable “black box.” Another crucial step is investing in explainability. This means using AI techniques that can provide a simple, human-readable reason for their outputs, so a frontline official can understand why a certain area was flagged as high-risk, for example.
An effective transparency framework is built on three pillars. First is auditing—creating mechanisms for independent parties to regularly review the models for bias and performance against real-world outcomes. Second is establishing clear lines of accountability, so it’s always understood who is responsible for a model’s output. And third is creating feedback loops, allowing officials and even the public to challenge or report anomalies in the system’s decisions. When you combine rigorous documentation, explainable outputs, and independent oversight, you build a system that can be trusted, scrutinized, and continuously improved.
To move from insight to impact, a key strategy is to blend official statistics with alternative data sources. Can you describe a scenario where this hybrid approach is crucial and explain the methods used to manage uncertainty while maintaining credibility with policymakers?
A perfect scenario is monitoring economic recovery after a major flood or earthquake. Official statistics, like business registration renewals or tax receipts, are highly accurate but can take months to compile. Policymakers on the ground can’t wait that long to make decisions about where to direct aid. A hybrid approach becomes crucial here. A statistical office could blend its baseline official data with high-frequency alternative sources, such as satellite imagery showing nighttime lights returning to a commercial district, or anonymized mobility data indicating a rise in retail footfall. This creates a “nowcast” of economic activity that is available weekly, not quarterly.
The key to maintaining credibility is to be completely transparent about the methods. Analysts must clearly document how the different sources are weighted and integrated. Crucially, they must quantify and communicate the uncertainty. The output isn’t presented as a definitive fact, but as a high-frequency estimate with a stated confidence interval. By framing it as a timely proxy designed to complement, not replace, the more rigorous official statistics that will follow, the agency preserves its reputation for accuracy while providing policymakers with the actionable, real-time insight they desperately need.
What is your forecast for the integration of AI and big data in achieving the Sustainable Development Goals over the next decade?
My forecast is one of pragmatic optimism. Over the next decade, I believe we will see a fundamental shift from using these technologies in isolated, headline-grabbing pilot projects to their systemic integration into the core operations of governance. The biggest change will be moving from using AI and big data simply for reporting on the SDGs to using them for the dynamic management of sustainable development. We’re going to build systems that don’t just tell us our deforestation rate last year, but that anticipate where illegal logging is likely to happen next week and help us intervene.
However, this future isn’t guaranteed by technology alone. Its realization will depend entirely on our investment in the human and institutional pillars: the skills of our public servants, the strength of our data governance, and the depth of our regional collaboration to share what works. The ultimate goal is to create feedback loops where policy is treated as an experiment, outcomes are monitored in real time, and our models and strategies are refined continuously. If we get this right, we will build systems where evidence doesn’t just sit in a report describing the world—it becomes an active force in changing it for the better.