Mapping Algorithmic Welfare Systems in India

India has witnessed an unprecedented proliferation of digitalisation, positioning itself as the third largest digitalised economy globally, surpassing technologically advanced nations such as Japan, the United Kingdom, and Germany. The country's digital economy has demonstrated considerable economic promise, with projections indicating an expansion to USD 1 trillion by 2028, driven by increased internet penetration and homegrown technological innovations such as UPI. 

Over the past decade, India's welfare architecture as a part of this digital transformation has undergone a quiet but profound shift. One of the core objectives of Digital India Mission, to provide government services digitally, has led to a structural transformation in the modes through which governance is organised and mediated.

Increasingly, public service delivery in India is mediated not by human discretion but by algorithmic systems, automated tools that sort, verify, and adjudicate citizen entitlements at scale. 

Our research on Algorithmic Welfare is examining how this shift is fundamentally restructuring the terms on which citizens access welfare services. Welfare was administered through physical verification, manual documentation, and field visits by government officials. These processes, at least in principle, were open to scrutiny and appeal. As the welfare delivery shifts to algorithmic systems, not only do millions of eligible claimants face losing these benefits, the process also lacks transparency leaving little room for accountability. 

This shift has reoriented the foundational presumption of welfare delivery. Rather than beginning from an assumption of entitlement, algorithmic systems are designed primarily to detect fraud, effectively treating claimants as suspects until the data proves otherwise. 

The concerns raised by existing algorithmic welfare systems take on renewed urgency in the context of India's accelerating embrace of artificial intelligence. Welfare algorithms currently rely on rule-based database matching, which identifies based on certain set criteria which ensures the technology can be tracked. However, AI-powered systems have far wider capabilities. Instead of following certain rules, they use machine learning to detect, make probability-based judgements, and continuously update their own decision making. These systems operate at a massive scale and speed making it structurally difficult to track error or biases.

To address how we can ensure technologies strengthen welfare, we are undertaking a project to map welfare algorithms currently in use across different Indian states. Through this project, we aim to analyse how eligibility decisions are made and where gaps in implementation may be leading to unintended exclusions. 

This project aims to generate accessible, evidence-based insights to inform and support policymakers, government departments, developers, and other key stakeholders to design responsible algorithmic welfare. By recognising recurring risk patterns, this project seeks to support responsible and accountable deployment of AI and automated decision making tools in welfare schemes, in order to strengthen public service delivery to the most marginalised. 

For more details read our explainer on Algorithmic Welfare Systems.