Samana Mehmood
A widowed lady labourer from rural Sindh, hurries to a cash-out shop to collect her BISP payment. The system shows the money is “in the system,” but the biometric reader flashes red: fingerprint mismatch. Dust, weak signals or worn fingertips block access. She returns home empty-handed, then comes back days later, paying transport twice and borrowing for groceries. The transfer meant to stabilize her household now destabilizes it. This is “algorithmic poverty”—poverty produced not just by economic conditions but by rules, databases and automated checks that determine who receives help.
Digitizing welfare is marketed as efficient and fraud-resistant. But the reality is harsher. Systems commit errors of inclusion and exclusion and the moral weight falls on exclusion, hitting those already at the edge. In India, stricter biometric ID rules in the food ration system reduced corruption but left 1.5–2 million eligible beneficiaries without access at some point. The lesson is not to reject digital systems, but to recognize that design, exception handling and grievance capacity determine whether technology reduces waste or creates deprivation.
Pakistan’s digital infrastructure is impressive, yet flawed. NADRA has issued around 120 million CNICs, but not “almost everyone.” Historically, children of single mothers, migrants, nomadic communities and many women were excluded. If CNIC possession is the gateway to welfare, digitization simply shifts exclusion upstream. Even with CNICs, identity must be continually “proven.” NADRA’s recent exploration of alternatives like Verisys and facial recognition reflects that “biometric or nothing” is a design choice, not a technological necessity.
BISP’s biometric verification, introduced in 2017, increased women’s direct collection and maintained average access. Yet early results showed withdrawal in a single attempt became significantly harder. Average access masks concentrated burdens on those least able to cope—manual workers, the elderly, the disabled and the chronically ill. Automated welfare is dangerous because it dissolves responsibility: a blocked payment becomes a “technical glitch” rather than a policy decision. Appeals exist but depend on time, literacy, transport and confidence, rendering the “right to appeal” costly and humiliating for the poor.
Procurement matters too. Vendors design scoring models, payment rails and complaint portals, but opacity shields them from accountability. Limited disclosure weakens democratic oversight and erects barriers to challenge.
Digital welfare can be inclusive, but only if systems are designed for it. Analogue fallbacks must work in the real economy, with alternative verification, supervised overrides and time-bound escalation. Exclusion must be published, including failed authentications, repeat visits, delays and appeal outcomes. No automated score or biometric failure should be final without rapid human review.
Outreach to remote areas, mobile units and low-cost procedures must correct errors without repeated travel. Identity and welfare must be rights, not benevolent favours. Algorithmic poverty is avoidable if exclusion stops being treated as a “glitch” and is recognized as public policy.
—The author is an MPhil scholar in Development Studies at the Pakistan Institute of Development Economics (PIDE). ([email protected])
