● Research Program
SARAL — Algorithmic Recommendations and Field-Level Discretion
A controlled research instrument for studying how administrative decision-makers respond to field-generated context when reviewing rule-based eligibility recommendations in welfare verification.
What SARAL is
SARAL is a session-based experimental instrument developed to examine a specific empirical question: when reviewers have access to a formal eligibility record, an algorithmic recommendation, and a field note from prior verification, how do they integrate these inputs into a final decision?
The work is grounded in the Slum Rehabilitation Authority (SRA) Annexure-II decision context in Mumbai, governed by the 2018 Government Resolution, with the primary experimental arm deployed online via Prolific.
Why it exists
Algorithmic decision-support systems in welfare administration are increasingly deployed to reduce reviewer burden. But formal eligibility records cannot capture every signal a field officer encounters.
The intended contribution is a portable research instrument for studying how field-generated context changes human responses to rule-based algorithmic recommendations — designed to be adapted across welfare contexts beyond its SRA grounding.
Phases
A two-phase research arc
A note on accountability
SARAL is conducted as independent research outside the formal requirements of any thesis. A self-governance framework documents participant protection, data handling, and research integrity in lieu of institutional ethics review. Pre-registration is timestamped before fieldwork; the pre-registration amendment is timestamped before Prolific data collection. All materials are publicly accessible.