Digital Identity, Exclusion and What It Means to Design Public Systems Well

A ration dealer in rural Jharkhand once explained to a visiting researcher why an elderly widow had gone without her monthly grain entitlement for three consecutive months: her fingerprints, worn thin by decades of manual labour, no longer matched the biometric template captured years earlier when her Aadhaar card was issued. The system was not malfunctioning in any technical sense; it was working exactly as designed, verifying identity through fingerprint matching with a threshold tolerance that did not account for the physical reality of an ageing agricultural labourer's hands. Her exclusion was not an accident. It was, in a strict sense, a design decision, made by engineers and policymakers who had to choose some threshold and chose one that worked for most people most of the time.

The scale of what Aadhaar solved

It is worth stating plainly what Aadhaar achieved before examining what it got wrong, because critics of the programme sometimes elide the genuine problem it addressed. Before Aadhaar, a significant share of Indians, particularly the poor, migrants and women, had no reliable, portable proof of identity accepted uniformly across states and departments, which meant they were routinely excluded from banking, welfare and formal employment for want of documentation rather than for want of eligibility. Aadhaar's biometric and demographic database, now covering well over a billion residents, gave India something close to a universal identity layer, and studies by researchers including those affiliated with the erstwhile Unique Identification Authority of India have documented meaningful reductions in leakage for schemes like the Public Distribution System and MGNREGA wages, where ghost beneficiaries and duplicate entries had previously diverted a nontrivial share of subsidy outlays.

Where authentication meets human reality

The trouble is that a system built for identification, proving that a person exists and is who she claims to be, was extended into authentication for time-sensitive welfare delivery, verifying that this specific person is present at this specific moment to receive this specific benefit, and that extension exposed a set of failure modes that identification alone never had to confront. Manual labourers, the elderly, and people with certain skin conditions or occupational hand injuries frequently produce fingerprints that biometric scanners reject, a well-documented pattern in research conducted by the Centre for Internet and Society and others. Iris scanning, offered as an alternative, requires equipment and connectivity that many of India's remotest ration shops and Aadhaar-enabled payment points simply do not reliably have, particularly in hill and forest regions with patchy electricity, let alone broadband.

When the stakes are survival

These are not abstract inconveniences. Multiple documented cases, including deaths attributed by activists and some state government inquiries to starvation following repeated denial of ration entitlements due to authentication failure, most notably a cluster of cases in Jharkhand between 2017 and 2019 that prompted the state government itself to review its biometric mandate, illustrate what happens when a system's edge-case failure rate, however statistically small, intersects with populations that have no buffer against a missed month of subsidised grain. The Supreme Court's 2018 Puttaswamy judgment upholding Aadhaar's constitutionality while striking down its mandatory linkage to bank accounts and mobile numbers implicitly recognised this tension, insisting that authentication failure could never be grounds for denial of a legal entitlement and that alternative verification must always remain available. The gap between that legal principle and its implementation on the ground in ration shops across the country remains real, though it has narrowed since states began mandating manual override options more consistently.

The design lesson that generalises

The deeper lesson here extends well beyond Aadhaar to any public digital system India builds going forward, including the health stack under Ayushman Bharat Digital Mission and the education credentialing infrastructure under ABC, the Academic Bank of Credits. A system's average performance, the percentage of transactions that succeed smoothly, is the wrong metric by which to judge public infrastructure that mediates access to survival-critical entitlements. The right metric is failure distribution: who fails, how often, and what recourse exists when they do. A ninety-nine percent success rate sounds impressive in a press release, but if the excluded one percent is disproportionately composed of manual labourers, the elderly, and people with disabilities, precisely the populations welfare systems exist to serve, then the system has reproduced through code the same exclusions it was meant to eliminate through documentation.

Building in graceful failure

Designing for this reality means building graceful degradation into public systems from the outset rather than retrofitting it after activist pressure and court orders. It means multiple authentication pathways running in genuine parallel rather than as a bureaucratically discouraged fallback, offline modes that function where connectivity does not reach, and grievance redress mechanisms that operate on a timeline compatible with the urgency of the entitlement at stake, since a grievance filed against a denied ration transaction that takes three weeks to resolve defeats its own purpose. Some of this has improved; the government's own periodic reviews and NITI Aayog's engagement with state-level exclusion data suggest a system that has learned, however unevenly, from its early failures.

None of this is an argument against digital identity infrastructure, which India needed and largely built well by global standards of ambition and scale. It is an argument that the legitimacy of such infrastructure rests entirely on how it treats the person for whom the system does not work as intended, because that person, by definition, cannot advocate for herself within a system that has already failed to recognise her. A country that wants to be judged by the sophistication of its digital public infrastructure must also be willing to be judged by how it treats the residual few percent that infrastructure was never quite built to see.