Part 4 · The Missing Signal
The Missing Signal
This part considers why app-first welfare systems often fail for basic-phone users and low-literacy households, and what more inclusive intake might look like. It also reflects on where algorithmic systems may help, where they may create new risks, and how SARAL was imagined as a transparency layer rather than a replacement for people.
India’s welfare systems went digital with a promise of simplification. On the ground, many people spoke less about algorithms and more about apps they could not open, forms they could not read, and the operators who made the process possible.
The myth of the app-first citizen
Many state portals assume smartphone ownership and confidence with user interfaces. In practice, respondents often described shared devices, basic phones, or reliance on an employer’s phone. In that setting, an app-first welfare system can feel less universal than it appears on paper.
“I applied online” often meant a nephew uploaded someone’s Aadhaar. When that application was rejected, the person who needed the benefit had no easy way to check status or appeal.
Construction worker, MumbaiWhen algorithmic systems help — and when they harm
Rule-based screening and automated checks can speed verification and reduce some forms of arbitrary discretion. At the same time, such systems may also absorb historical patterns unless someone actively audits and questions them. If past approvals cluster by district, caste, or occupation, a system can quietly learn those patterns as normal.
“At least when the officer says no, I can ask why. The computer doesn’t talk.”
Respondent, MaharashtraThe missing ingredient: transparency
Trust rarely comes from full automation in context-rich welfare settings. It appears more likely to come from understanding. When people can see how decisions are being made, and when operators can explain what happened, confidence begins to rebuild.
How SARAL was imagined differently
Applicants would receive plain-language outcomes such as:
- “Your application was paused because income proof was missing.”
- “To appeal, send ‘APPEAL’ to 720XXX within 7 days.”
For operators, the system would show:
- Which rules fired, what flags were raised, and where discretion applied.
- Where judgment was used, so teams could reflect on patterns of overrides and procedural friction.
Designing for inclusion, not perfection
Inclusive intake begins with listening. One recurring lesson from SmartAgro through Project Nagrik was that simple channels, voice prompts, SMS check-ins, and human verification counters, can matter more than interface sophistication because they preserve participation.
“If the machine listens, I don’t mind if it decides.”
Respondent, MumbaiFrom insight to action
The practical question is not machines versus humans, but which systems earn trust. A system that hides behind algorithmic opacity creates distance; one that explains itself may make public service feel more legible and contestable.
A bias-aware, citizen-visible direction
- SMS/IVR intake for wider accessibility
- Rules mapping that explains every decision
- Anomaly checks to surface potential bias
- Audit trails linking citizen-facing and operator-facing actions
This was the direction that informed SARAL: a transparency-oriented system intended to make access more understandable rather than simply more automated.
This closes the first arc of the blog series.
The posts above attempt to capture the everyday frictions citizens described while navigating welfare systems. Those observations shaped the early thinking behind SARAL v1, a system concept oriented toward transparency and more inclusive access.
The project continues to evolve through design refinement and later-stage testing under a formal research pathway.
Engineer & Data Scientist — Building AI for Governance