This paper studies how frontline operators interact with algorithmic eligibility signals in welfare triage settings. It is motivated by an initial field deployment of a decision-support system (SARAL) in India, which surfaced recurring observations about how operators engage with structured rule outputs alongside unstructured case information. These observations indicate potential mechanisms through which decision-making may be influenced by signals not encoded within formal rule systems, including material-affluence proxies, contextual spillovers, normative heuristics, and administrative constraints.

These observations were collected during an exploratory deployment and are not used for formal statistical or causal claims. To rigorously evaluate these mechanisms, the paper proposes a controlled experimental framework using synthetic profiles and a dual-assessor structure. This design isolates the causal effect of unencoded signals on decision outcomes and distinguishes between idiosyncratic operator judgment and institutionally aligned behavior.

The study aims to formalize the gap between algorithmic rule representation and real-world administrative decision-making under discretion.