VulnerabilityMap: An Open Framework for Mapping Vulnerability among Urban Disadvantaged Populations in the United States

Lin Chen, Yong Li, Pan Hui

Abstract

Cities are crucibles of numerous opportunities, but also hotbeds of inequality. The plight of disadvantaged populations who are "left behind" within urban environments has been an increasingly pressing concern, which poses substantial threats to the realization of the UN SDG agenda. However, a comprehensive framework for studying this urban dilemma is currently absent, preventing researchers from developing AI models for social good prediction and intervention. To fill this gap, we construct VulnerabilityMap, a framework to meticulously dissect the challenges faced by urban disadvantaged populations, unraveling their vulnerability to a spectrum of shocks and stresses that are categorized through the prism of Maslow's hierarchy of needs. Specifically, we systematically collect large-scale multi-sourced census and web-based data covering more than 328 million people in the United States regarding demographic features, neighborhood environments, offline mobility behaviors, and online social connections. These features are further related to vulnerability outcomes from short-term shocks such as COVID-19 and long-term physiological, social, and self-actualization stresses. Leveraging our framework, we construct machine learning models that exhibit strong performance in predicting vulnerability outcomes from various disadvantage features, which shows the promising utility of our framework to support targeted AI models. Moreover, we provide model-based explainability analysis to interpret the reasons underlying model predictions, shedding light on intricate social factors that trap certain populations inside vulnerable situations. Our constructed dataset is publicly available at https://github.com/LinChen-65/VulnerabilityMap/.

Introduction

While cities are often celebrated as engines of innovation and economic growth, they simultaneously function as "hotbeds of inequality." Current research into urban disadvantage is frequently siloed, focusing on isolated metrics such as income or race. This fragmented perspective hinders the development of effective AI-driven interventions for social good. Vulnerability is a complex "web" of interconnected factors—demographics, neighborhood environment, offline mobility, and online social connections. To address this, VulnerabilityMap provides a unified framework that relates multi-sourced data to outcomes ranging from physiological health to self-actualization.

The VulnerabilityMap Framework

The framework is structured around two primary components: Disadvantage Features and Vulnerability Outcomes. Disadvantage is measured across four dimensions: Demographic, Neighborhood, Mobility, and Social. Outcomes are categorized into short-term "shocks" (e.g., COVID-19) and long-term "stresses." Borrowing from Maslow’s hierarchy of needs, these stresses are further classified into Physiological, Social, and Self-Actualization categories.

Overview of the VulnerabilityMap framework, detailing the specific features used to construct disadvantage indices and the taxonomy of vulnerability outcomes.

Methodology and Data Integration

The researchers integrated large-scale data covering over 328 million people in the US. This includes American Community Survey (ACS) census data, SafeGraph mobility patterns, Facebook social connectivity and interest data, and Twitter sentiment indices.

The workflow involves a three-step process to create a Composite Disadvantage Index:

  1. Ranking: Neighborhoods are ranked by disadvantage across the four dimensions.
  2. Normalization: Applying z-score normalization to ensure comparability across disparate data types.
  3. Aggregation: Summing these normalized values to form the final composite index.
The VulnerabilityMap workflow, illustrating the path from multi-sourced raw data to the construction of disadvantage indices and the prediction of vulnerability outcomes.

Predictive modeling was performed using Random Forest regression, chosen for its high performance and explainability. This allows for the extraction of feature importance scores, revealing the primary drivers behind specific vulnerabilities.

Key Insights and Results

The framework's predictive models achieved high accuracy, with 15 out of 20 outcomes reaching an $R^2 > 0.7$. Beyond prediction, the framework yielded several critical social insights:

  • The "Support Ratio" Paradox: Demographic disadvantage often correlates with higher "support ratios." While these neighborhoods lack broad resources, they maintain tight-knit social networks, likely because limited mobility confines residents to local family and friend circles.
  • Education vs. Arts: Access to education is primarily driven by social connections ("who you know"), whereas engagement with arts and culture is more closely tied to education levels ("what you know"). This suggests that aesthetic training is less transferable through social networks than general information about schools.
  • Mobility as a Proxy for Choice: During COVID-19 lockdowns, the link between demographic status and mobility weakened. Wealthier populations could choose to stay home, while disadvantaged populations—often essential workers—were forced to maintain mobility, increasing their vulnerability to the shock.

Conclusion

VulnerabilityMap successfully identifies macroscopic trends, such as high-disadvantage clusters in the Southwest and low-disadvantage hubs in the Northeast. While the framework currently faces challenges regarding data latency and simple aggregation methods, it represents a significant step toward targeted urban planning. Future work will focus on developing a public web interface to assist policymakers in addressing urban inequality through data-driven interventions.

Summary updated

Abstract

Cities are crucibles of numerous opportunities, but also hotbeds of inequality. The plight of disadvantaged populations who are ``left behind'' within urban environments has been an increasingly pressing concern, which poses substantial threats to the realization of the UN SDG agenda. However, a comprehensive framework for studying this urban dilemma is currently absent, preventing researchers from developing AI models for social good prediction and intervention. To fill this gap, we construct VulnerabilityMap, a framework to meticulously dissect the challenges faced by urban disadvantaged populations, unraveling their vulnerability to a spectrum of shocks and stresses that are categorized through the prism of Maslow's hierarchy of needs. Specifically, we systematically collect large-scale multi-sourced census and web-based data covering more than 328 million people in the United States regarding demographic features, neighborhood environments, offline mobility behaviors, and online social connections. These features are further related to vulnerability outcomes from short-term shocks such as COVID-19 and long-term physiological, social, and self-actualization stresses. Leveraging our framework, we construct machine learning models that exhibit strong performance in predicting vulnerability outcomes from various disadvantage features, which shows the promising utility of our framework to support targeted AI models. Moreover, we provide model-based explainability analysis to interpret the reasons underlying model predictions, shedding light on intricate social factors that trap certain populations inside vulnerable situations. Our constructed dataset is publicly available at https://github.com/LinChen-65/VulnerabilityMap/.
BibTeX
@inproceedings{chen2024vulnerabilitymap,
  author = {Lin Chen and Yong Li and Pan Hui},
  title = {{VulnerabilityMap: An Open Framework for Mapping Vulnerability among Urban Disadvantaged Populations in the United States}},
  booktitle = {International Joint Conference on Artificial Intelligence},
  year = {2024},
  url = {https://openreview.net/forum?id=5SyJWeUcrt},
}
Figure