AI-Assistance for Resilience Research and Practice

Summary

The Community Resilience Program has traditionally focused on creating planning guidance, resilience metrics, and integrated models at the community scale. These tools have empowered stakeholders to assess the impacts of hazard events, prioritize solutions, and support decision-making for improved community resilience. Previous efforts have successfully embedded resilience concepts into guidance, standards, and code documents through collaboration with professional organizations and agencies.

Developing such tools and resources is time-consuming and labor intensive. Moreover, despite the development of robust tools, resilience professionals, particularly within the federal government, often face challenges in providing adequate technical support, impeding community uptake, especially among those lacking technical expertise. This issue is more pronounced in underserved communities.

The primary goals of this project are to:

1. Assist resilience professionals with technical tasks by leveraging artificial intelligence (AI) and human-in-the-loop approaches to streamline the development of resilience tools and resources; and
2. Bridge the technical gap between the tools produced by resilience professionals and the end-users, such as state and local governments, to better support resilience efforts.

By using AI and human-in-the-loop methodologies, the project aims to enhance the efficiency and effectiveness of developing resilience tools, while also making these tools more accessible and actionable for a broader range of users, ultimately improving community resilience outcomes.

Description

Objective - In FY25, the project will focus on developing and evaluating human-in-the-loop AI-assisted methods for data curation, data analysis, and behavioral experiments. These efforts will support decision-making to enhance community resilience against compound risks from extreme weather events and other threats by providing more accessible, reliable, and actionable data. Integrating human expertise at critical stages ensures trust and reliability in these AI applications. This initiative aligns with EL’s Strategic Goal of creating Disaster-Resilient Buildings, Infrastructure, and Communities.

By 2026, the project aims to:

1. Develop and validate at least three AI and natural language processing (NLP) methodologies, emphasizing human oversight, to assist in analyzing technical texts related to resilience planning.

2. Implement and refine at least two human-in-the-loop AI-assisted methods to improve data curation, analysis, and decision-making for community resilience.

What is the technical idea? 

AI and NLP methodologies can assist in analyzing complex technical texts related to resilience, which are often challenging for non-experts to comprehend. By tailoring these methodologies specifically for resilience-related texts, the project aims to facilitate the extraction of valuable insights from diverse data sources such as community planning documents, social media data, and scientific literature.

What is the research plan? 

The research plan is centered on human oversight in (1) developing and validating AI-assisted methods and (2) advancing these methods to practical applications in community resilience. This iterative approach involves continuous testing, validation, and refinement, ensuring that AI methodologies remain reliable and relevant. In the long term, there is an opportunity to develop best practices or standards for validating AI-assisted human-in-the-loop applications for technical tasks.

Core Components:

1. Human-in-the-Loop AI-Assisted Methods:
   1. Develop and validate AI and NLP methodologies for analyzing technical texts on resilience planning.
   2. Conduct rigorous testing and validation with strong human oversight.
2. Human-in-the-Loop AI-Assisted Applications:
   1. Enhance AI-assisted methods for data curation, analysis, and decision-making in community resilience.
   2. Implement strategies to validate AI outputs through collaboration with domain experts and community stakeholders.