The humanitarian sector is at the pinnacle of a technological revolution that can fundamentally change the way we serve the world's most vulnerable groups. New research paper from the University of London – Interaction between artificial intelligence and digital public infrastructure: concepts, benefits, challenges – Avoid how artificial intelligence and digital public infrastructure (DPI) can work together to create more comprehensive, efficient and equitable public services.
Definition of a game changer
- AI is a general purpose technology (GPT) like electricity with a wide range of applications across society.
- DPI refers to the underlying digital system that enables everyday functions such as digital payments, identification, and data exchange, such as the Aadhaar Digital ID system in India.
Magic happens when these technologies intersect. The versatility of AI means that virtually every DPI system can be enhanced, and DPI provides structured, consent-based data to make AI systems more accurate and comprehensive.
Real-world impact
The most compelling evidence comes from countries that have already implemented these combinations. India's Basini system exemplifies the potential for change. This AI-powered translation platform uses natural language processing, the most commonly used machine learning technique in texts, to perform rapid translation between Indian languages. The system utilizes crowdsourcing collaborators across the country to improve translation, particularly in languages with limited data.
For humanitarian organizations working in multilingual contexts, this represents a breakthrough. Imagine being able to provide high-quality translation services in real time in dozens of local languages.
Singapore offers another powerful example of SingPass. It recently integrated machine learning for user authentication to combat fraud. As research has pointed out, this can help improve public trust in the system and promote broader uptake.
For humanitarian programs that deal with the challenges of identity verification, such AI-enhanced authentication can dramatically improve both security and user experience.
The Danish Muni chatbot, which serves 37 municipalities, shows how AI can streamline service delivery. This refers to a future in which displaced groups can interact with AI assistants in their own language to navigate all the complex bureaucratic processes, from asylum applications to healthcare access.
Data Foundation Revolution
Perhaps even more important is how DPI acts as the foundation for a more comprehensive AI system. Researchers highlight important issues. “Large language models will consume data generated by all humans by 2028.” Synthesis alternatives prove insufficient.
DPI provides solutions through large-scale consent-based data collection. India's Aadhaar system, with nearly 1.38 billion users, demonstrates this possibility. The Indian government has created a public dataset from this data that AI startups can use to develop better systems. This is essentially an unofficial version of AI's “industrial policy.”
DPI helps to address algorithm bias. This study explains that certain historically marginalized populations are often underestimated in AI datasets and lead to algorithmic bias. However, when DPI systems achieve universal adoption, they can capture data from marginalized communities, including traditional knowledge of Indigenous peoples, making AI systems more representative and accurate for these groups.
Face the challenge
Researchers don't sugar coat obstacles. High inference costs raise immediate concern. If DPI serves hundreds of millions or billions of citizens, AI integration could put a strain on government budgets. The challenges of interoperability with legacy systems create additional barriers, particularly when outdated technologies are associated with common development contexts.
Ethical considerations are more concerning.
The advantages of using DPI data for AI development involve either a comprehensive and effective implementation of DPI or the risk of becoming a double-edged sword. If marginalized communities do not adopt the DPI system due to systemic differences or distrust, the resulting dataset can actually exacerbate algorithm bias.
Privacy represents another important issue. As researchers emphasize, it is important that all data are collected with individual informed consent before collection. As such, DPIs run the risk of infringing on the rights of individuals that are important to privacy.
Development orders
For humanitarian organizations, these developments require immediate attention. Countries that implement AI-enhanced DPI will jump over others to service delivery capabilities. Organizations that understand and advocate for comprehensive AI-DPI integration are better positioned to effectively serve beneficiaries.
We need to promote implementation of DPIs that prioritize universal access, ensuring marginalized groups are not left behind. We need to advocate for transparent, ethical data governance that respects privacy while enabling innovation. Most importantly, we must start experimenting with AI-enhanced service delivery with our own programs.
The convergence of AI and DPI is not just a technical trend, but a humanitarian order. Rightly, it could democratize access to high quality public services worldwide. If you're wrong, it can deepen existing inequality. The choice is ours, but the windows for influence are rapidly narrowing.
