This research introduces a novel approach for interactive decision-making agents that can determine eligibility across complex domains without requiring manual annotation or decision trees.

• Combines program synthesis with dialogue capabilities to create agents that can make binary decisions based on user features
• Enables automated assistance for complex domains like legal codes and medical diagnosis
• Outperforms traditional methods by maintaining conversation context while converting natural language into executable programs
• Demonstrates practical applications for navigating complex eligibility requirements in frequently updated legal frameworks

For legal professionals, this technology offers a powerful way to automate eligibility determination for social benefits, legal compliance checks, and regulatory navigation without constant manual updates to systems.

Program Synthesis Dialog Agents for Interactive Decision-Making