Agentic AI for Child Welfare Intake and Early Risk Identification: A Governance-Aware Framework

Area of Science:

• Computer Science
• Artificial Intelligence
• Social Sciences

Background:

• Child welfare intake is a critical but often inconsistent process.
• Missed details and individual judgment can delay aid for vulnerable children.
• Existing systems lack robust governance and bias monitoring.

Purpose of the Study:

• To propose a governance-aware AI framework for child welfare intake.
• To develop a multi-agent AI system for early risk identification.
• To enhance the consistency, completeness, and governance of child welfare intake processes.

Main Methods:

• Developed a multi-agent AI system with a conversational chatbot interface.
• Implemented agents for data analysis, risk assessment, quality checks, and bias monitoring.
• Designed a Central Orchestrator to manage agents and workflow, including human-in-the-loop protocols.

Main Results:

• The AI system supports early risk identification and data analysis.
• It incorporates human-in-the-loop mechanisms for uncertain or high-risk cases.
• The system aims to improve data quality and reduce bias in intake decisions.

Conclusions:

• The proposed AI system can enhance child welfare intake processes.
• It offers a supportive tool for caseworkers, not a replacement.
• With validation, it can help vulnerable children receive timely attention.