Roadmap Towards Responsible AI in Crisis Resilience Management

AI in crisis management requires a holistic approach to responsibility: The paper presents six propositions embedded in a "Crisis Information Management Circle," indicating that responsible AI cannot be addressed in isolation but must be integrated throughout the entire lifecycle of crisis response—from data collection through analysis, sharing, and decision-making.

Vulnerable populations face disproportionate risks from biased AI systems: The roadmap specifically emphasizes the need to protect vulnerable populations, suggesting that AI systems used in crises can amplify existing inequalities if equity and fairness considerations are not deliberately built into their design and deployment.

Multiple stakeholders must coordinate on responsible AI standards: The paper identifies researchers, practitioners, developers, emergency managers, humanitarian organizations, and public officials as key audiences, indicating that establishing responsible AI practices in crisis management requires coordination across traditionally siloed sectors rather than solutions developed by any single group.

Transparency and explainability are critical for crisis decision-making: The roadmap addresses explainability and transparency as core challenges, suggesting that during crises—when decisions directly impact public safety—AI systems must be understandable to decision-makers and the public, not treated as "black boxes."

Privacy, security, and information sharing create competing tensions: The paper identifies privacy and security alongside the need for inter-organizational data sharing, highlighting a fundamental challenge: crisis response often requires rapid information exchange, yet this conflicts with protecting individual privacy and preventing misuse of sensitive data.