Skip to content

Benefits Automation: From Rules as Code to Caregiver Discovery

An estimated $60 billion in benefits go unclaimed every year5. This page synthesizes research on automating benefits eligibility through Rules as Code and knowledge graphs, supporting caregiver-specific assessment and benefit discovery.

Rules as Code: Translating policy into executable logic

The Beeck Center at Georgetown University documented four experiments translating SNAP and Medicaid policy into executable rules using large language models2. The research showed that LLMs can support translating policy into code, but require external knowledge and human oversight within an iterative process for policies containing complex logic2.

These experiments do not establish GiveCare's screening accuracy. Any adopted rules need evidence for the program, jurisdiction, and period they cover.

What Rules as Code does not solve

Rules as Code translates individual program rules into executable logic. It does not:

  • Model interactions between programs (how enrolling in one affects eligibility for another)
  • Represent caregiver-specific eligibility pathways (most rules are written for the benefit recipient, not their caregiver)
  • Handle benefit cliffs (where small income changes trigger large benefit losses)
  • Connect eligibility to discovery (knowing someone qualifies is useless if they do not know the program exists)

These gaps require additional infrastructure.

Knowledge graphs: Modeling program interdependencies

Knowledge graphs and ontologies can model the complex relationships between health and social programs, enabling reasoning about how enrolling in one program affects eligibility for others3. Ensuring consistent application of policy rules at population scale is critical to ensuring that scarce resources get to those in need3.

Where Rules as Code handles individual program logic, knowledge graphs represent the connections between programs. Program interactions need their own source-backed rules. Medicaid enrollment alone does not establish eligibility for a waiver service. Graph-based representations can model income thresholds where benefits drop off across multiple programs simultaneously. Evaluating a household's benefit changes would require current rules for every affected program and the relevant household facts. GiveCare does not currently perform that calculation.

HCBS Taxonomy: The classification standard

The CMS Home and Community-Based Services taxonomy organizes services into 18 categories and over 60 specific services4 — the standard classification for services delivered to Medicaid beneficiaries in home and community settings.

Before this taxonomy, no consistent language existed for HCBS services across states. The same service had different names, different descriptions, and different scopes in every state Medicaid program. The taxonomy provides a reference for service classification across jurisdictions.

The gap: Unclaimed benefits and benefit cliffs

An estimated $60 billion in benefits go unclaimed every year5. Nearly one in four public assistance recipients report taking negative actions like declining raises or working fewer hours to avoid benefit cliffs6. 62 percent of recipients surveyed felt stuck in low-income situations due to concerns that higher earnings would trigger benefit loss6.

Why the gap persists

The $60 billion gap exists because three problems compound:

  1. Discovery failure. People do not know programs exist. GiveCare's verified owner-projected program directory is the current public inventory; this evidence page does not copy its count. No single source maps all programs. Discovery requires knowing what to search for, which requires knowing the system — a circular dependency.

  2. Navigation failure. Even when a program is discovered, applying requires navigating bureaucratic processes optimized for the system, not the user. Documentation requirements, application forms, interview scheduling, and verification procedures are designed for administrative efficiency, not user accessibility.

  3. Cliff avoidance. The 22% who take negative actions to avoid benefit cliffs6 are making rational decisions given the information available to them. Without visibility into how income changes affect the full portfolio of benefits, avoiding the cliff is safer than risking a gain that triggers a larger loss.

What the research does not yet combine

Individual components of the solution exist:

Component Exists in Gap
Rules as Code Beeck Center experiments2 Program-by-program, not cross-program
Knowledge graphs IBM Research3 Enterprise/government context, not user-facing
Service taxonomy CMS HCBS4 Classification, not discovery or navigation
Participation estimate Code for America5 Estimates the gap, does not measure GiveCare
Cliff analysis AEI6 Policy analysis, not user-facing tool

The evidence suggests a larger design target:

  • Caregiver SDOH assessment (understanding the caregiver's situation across all six domains)
  • Cross-program screening (checking a reviewed program bundle without claiming eligibility)
  • Cliff-aware portfolio analysis (modeling how changes affect the full benefit portfolio)
  • AI-guided discovery and navigation (presenting programs one at a time with clear next steps via SMS)

GiveCare currently connects domain-based screening, a verified owner-projected program directory, and cautious benefits discovery. Mira retrieves fresh candidates, checks their relevance, and applies deterministic prescreening to stated facts1. Each factual reply sentence must cite an observed passage. Cross-program optimization and benefit-cliff calculations remain research directions.

Research-informed direction

The research points to possible later layers:

  1. Rules as Code for individual program eligibility (Beeck Center validated approach)
  2. Knowledge graph for cross-program interactions and cliff detection (IBM validated approach)
  3. HCBS taxonomy for service classification and cross-state mapping (CMS standard)
  4. Caregiver SDOH as the input layer that no other system provides

The research supports each component in its own setting. It does not prove that GiveCare has integrated or validated the full stack. Any later layer must earn its own evidence and enter consumers through a verified owner projection.

  1. GiveCare. "GiveCare SMS Runtime Contract." Source → ↩

  2. Beeck Center at Georgetown University. "AI-Powered Rules as Code." 2025. Source → ↩↩↩

  3. IBM Research. "Knowledge Graphs for Social Good: Protecting Vital Health and Social Programs." 2021. Source → ↩↩↩

  4. Peebles, V. & Bohl, A. "The HCBS Taxonomy: A New Language for Classifying Home- and Community-Based Services." 2014. Source → ↩↩

  5. Code for America. "Making Our Systems See People." Source → ↩↩↩

  6. American Enterprise Institute. "Stranded by the Safety Net: How to Fix the Benefit Cliff Problem." 2024. Source → ↩↩↩↩