Built To Include, Designed To Exclude

In July 2025, the Ministry of Women and Child Development made facial recognition mandatory for pregnant and lactating women claiming take-home rations under POSHAN Abhiyaan. On the eve of the rollout, Anganwadi workers across the country were instructed to reinstall the POSHAN Tracker app.

The updated version had quietly stripped out the option to skip a failed scan. What had been framed as a modernisation of welfare delivery had, in a single software update, become a new condition for accessing it.

The reporting that followed, from Decode Internet and others, has documented what predictably happened next. Women turned away because a live scan would not match an Aadhaar photograph taken years earlier, sometimes before pregnancy changed their features, often in lighting and image quality that were never good enough to begin with.

Workers spending twenty minutes per right-holder on authentication, eating into hours meant for preschool teaching and nutrition counselling. Centres in rural Karnataka, Assam and Jharkhand where the app freezes, the connectivity drops, and the queue under a neem tree gets longer.

The question worth sitting with is not just how this happened, but why a technology so obviously unsuited to the task was mandated anyway, at scale, for some of the most vulnerable people in the country.

Solving The Wrong Problem

Facial recognition was framed as an anti-leakage measure. But the leakage in supplementary nutrition is not, in the main, an identity problem. It is a quantity problem: rations diluted, weights short, distributions skipped.

A live face match cannot tell you whether the bag a woman receives contains what it is supposed to contain. The technology cannot detect the fraud it was deployed to address. It can, however, very efficiently turn entitlement into a conditional privilege.

The answer, I think, has less to do with any specific policy failure and more to do with what digitisation has come to mean in Indian public administration. It has become ideological — a signal of modernity, of efficiency, of a state that has presumably moved beyond the messiness of paper records and human discretion.

Whether it actually works for the people it is supposed to serve has become, in a quiet but consequential way, secondary to whether it exists and can be counted. Enrolment figures rise, coverage expands, dashboards fill up. The woman who was turned away, the worker whose wages did not come through, the application that has been pending for three years — none of this shows up in any official account of how the system is performing.

The pattern repeats across the welfare architecture. MNREGA, which will be formally repealed on July 1 2026 and replaced by the Viksit Bharat–G RAM G Act, 2025, spent its final years tangled in exactly these contradictions.

Wages flowed through Aadhaar-based payments that depended on a long chain of conditions: correct seeding, active bank accounts, successful authentication. When any link in that chain broke, the worker had to navigate a redressal process that assumed time, literacy, and access to officials that most MNREGA workers do not have.

App-based attendance was its own quiet exclusion. A worker might be present on the muster roll for nine days; the app might show four; her wages reflected the app. Nobody disputed she was there. The system simply had no record of it — and the system was what counted.

The scheme is now ending. The questions it leaves behind, about who was actually paid and who fell through the authentication gaps, have never been seriously asked, let alone answered.

Built Around an Imagined User

What is striking about all of this is that the people most likely to be excluded by these requirements are precisely the people the schemes were designed for. Agricultural workers whose fingerprints have worn smooth from decades of physical labour and cannot authenticate biometrically. Women in remote areas who are quietly dropped from the right-holder lists when their attendance, now tracked digitally, lapses.

Migrant workers trying to access entitlements in a state they do not live in, through a portal in a language they do not read. These are not edge cases or implementation hiccups. They are predictable consequences of designing systems around an imagined user — connected, literate, fixed in place — who bears very little resemblance to the actual user these systems are meant to serve.

None of this is an argument against digital infrastructure in public services. India’s Digital Public Infrastructure is being held up internationally as a model worth copying.

Direct benefit transfer has meaningfully reduced leakage in some schemes, and there are genuine gains that should not be dismissed. But there is a difference between digitisation as a tool that serves delivery, and digitisation as a performance of modernity that substitutes for it.

The willingness to deploy facial recognition to solve a fraud problem it cannot actually solve, or to make OTP-based authentication mandatory for women whose Aadhaar-linked phone is in their husband’s pocket miles away, suggests we have drifted further toward the latter than we might acknowledge.

The fixes are not mysterious. Interfaces in local languages. Real offline alternatives — not the “skip” button that was just removed — for places where biometric or connectivity failures are foreseeable. Grievance mechanisms with actual people behind them, not portals that say “already registered” and offer no next step.

These get treated as secondary concerns, to be retrofitted later once problems surface upon rollout. For the people on the other side of a failed authentication or a frozen payment, they are not secondary at all. They are the difference between a system that works, and a system that has your name in it.

Counted, But Not Served

Prerna (name changed) is a domestic worker who works with us in Delhi and is originally from Bengal. She has been trying to get a ration card for nearly three years. Her daughter filled out the online form on her behalf, navigating a portal that, language options notwithstanding, assumes a level of digital familiarity neither of them really has.

The application has been stuck at “pending” ever since, with no indication of what is missing or what needs to be done. We tried submitting a fresh application using her Aadhaar details and OTP verification. The system simply said she was “already registered”.

That was it. No next step, no person to speak to, no way to resolve it.

In the meantime, she continues to spend significantly more on basic provisions than she should have to, because the portal has her on record but will not let her through, and there is no bell to ring.

She and the pregnant women turned away at the Anganwadi centre have never met, and their situations are different in almost every way. But they are caught in the same logic: a system that has been built to demonstrate that it includes people, and has not been built to actually serve them.

Until that distinction starts to matter as much as the enrolment numbers do, both of them will keep waiting.