Strengthening the Reliability, Resilience, and Integrity of Public-Benefit Information Systems: A Literature Review of Data-Driven Approaches in U.S. State Workforce and Public-Benefit Agencies
Main Article Content
Abstract
State workforce and public-benefit agencies administer the programs that millions of Americans rely on in moments of acute need, among them unemployment insurance (UI), the Supplemental Nutrition Assistance Program (SNAP), Temporary Assistance for Needy Families (TANF), and Medicaid. The information systems that deliver these services are often decades old, built for ordinary rather than crisis-level demand, and fragile under stress. The COVID-19 pandemic made this fragility plain. Claim volumes surged, aging systems faltered, and tens of billions of dollars were paid improperly even as eligible workers waited weeks for assistance. This paper reviews the governmental, scholarly, and technical literature on the reliability, resilience, and integrity of these systems and on the supervised and unsupervised data-mining methods proposed to support them. The review is organized around four themes: the fragility of legacy benefit infrastructure; resilience and cybersecurity; payment integrity and fraud; and the applied analytics used to address these problems, including demand forecasting, anomaly detection, and claimant segmentation. The literature points to a common lesson. Data-driven models offer real operational value, but that value is limited by predictive accuracy, by the uneven cost of false positives, by how well models hold up when conditions change, and by the procedural safeguards that surround their use. The paper closes with implications for treating reliability, resilience, and integrity as related design goals rather than separate ones.