Human-AI Use Patterns for Decision-Making in Disaster Scenarios: A Systematic Review

AI can enhance disaster response efficiency and situational awareness: The review of 51 studies found that Human-AI Decision Support Systems improve response efficiency and help decision-makers process complex information during high-stakes emergencies, though the paper notes these benefits depend on proper system design and implementation.

Explainability and transparency are critical but often lacking: The review identifies "Explainable AI" as a key category for building trust, yet highlights that many current systems fall short on interpretability—meaning emergency responders may not understand why an AI system made a particular recommendation, which is problematic when lives are at stake.

Current AI systems struggle with real-world disaster complexity: The paper surfaces "critical limitations in scalability, interpretability, and system interoperability," suggesting that while AI shows promise in controlled settings, existing systems often cannot adapt to the unpredictable, dynamic nature of actual disasters or work seamlessly with other tools responders rely on.

Training and simulation can prepare humans to work effectively with AI: The review identifies "virtual training environments" as an important category, indicating that properly training disaster responders to collaborate with AI systems—rather than simply deploying AI—is essential for responsible implementation.

Equitable outcomes require context-aware, adaptive systems: The paper emphasizes the need for "context-aware Human-AI systems" to support "equitable recovery outcomes," suggesting that one-size-fits-all AI solutions risk leaving vulnerable populations behind unless systems account for local conditions and diverse community needs.