Skip to main content
Asia AI4DObservatory
Responsible AI Use Case

Human-in-the-loop review for welfare eligibility triage

A public agency introduced automated triage for social protection applications, paired with a mandatory caseworker review step and a published appeals route.

Sri LankaPublic SectorGovernmentHuman OversightTransparencyAccountability

Illustrative prototype profile; it does not represent a verified deployment.

View related research

Overview

Placeholder overview content written at realistic length so the reading layout can be evaluated properly. The initiative applies a rules-and-model hybrid to order incoming social protection applications by likely eligibility, with every negative outcome routed to a human caseworker before any decision is communicated to an applicant.

Problem / Context

Placeholder context describing application backlogs, uneven processing times across districts, and the equity concerns raised by civil-society organisations during consultation.

A second paragraph would normally set out the legal and institutional environment, including the mandate of the implementing agency and any relevant data protection obligations.

AI Application

Placeholder description of the technical approach: what data is used, what the model predicts, where it sits in the workflow, and what it explicitly does not decide.

Responsible AI Considerations

Placeholder discussion of the safeguards adopted — documented model limitations, disparity monitoring across districts and household types, a published plain-language notice to applicants, and a named official accountable for the system.

This section is where the Observatory records both the safeguards claimed by the implementer and the gaps identified by independent reviewers.

Impact / Lessons

Placeholder lessons content. No impact figures are presented in this prototype; verified evaluation results will be added when supplied.

Why This Matters

Placeholder plain-language explanation aimed at non-specialist readers, connecting this initiative to the wider regional debate on automated decision-making in entitlement systems.