Reducing AI Model Biases with a Bilevel Learning Framework
A Case Study of Leveraging Twitter Data for Damage Estimation
Bibliographic Data
| ID | 3775569 |
|---|---|
| Authors | Weishan Bai (0009-0000-2006-4456, Urban Institute), Xinyue Ye (0000-0001-8838-9476, Urban Institute, corresponding author), Yiqun Xie (0000-0002-6439-1333, Geospatial Information Authority of Japan), Shannon Van Zandt (0000-0003-1640-7799, Technology Applications (United States)), Xiao Huang (0000-0002-4323-382X, Emory University), Debalina Sengupta (0000-0002-2026-2660, Energy Transitions (United Kingdom)) |
| Year | 2025 |
| Volume | 116 |
| Issue | 1 |
| Pages | 1-20 |
| Publication date | 2025-08-11 |
| Peer Reviewed | Yes |
| Open Access | No |
| Type | ARTICLE |
| Venue | Annals of the American Association of Geographers (JOURNAL) |
| Journal identifiers | ISSN: 2469-4452 • E-ISSN: 2469-4460 |
| Publisher | Informa UK Limited (PUBLISHER • GB) |
| DOI | 10.1080/24694452.2025.2536182 |
| OpenAlex | W4413219812 |
| Language | EN |
| References cited | 53 |
This study aims to improve disaster risk mitigation by integrating fairness into artificial intelligence (AI) models, specifically addressing spatial biases that can lead to unequal resource allocation during disasters. Our objective is to reduce spatial biases in disaster impact prediction models via a bilevel learning framework, enhancing both accuracy and fairness. To achieve this, we leverage information exchanges in social networks during the disaster period to estimate the economic loss caused by the disaster. Considering that accessible data are spatially biased in quantity and quality, this study applies a fairness framework within a deep neural network to mitigate spatial bias in predictions. We analyze Twitter data and the Federal Emergency Management Agency’s Real Property Damage Amount data to assess and predict the economic impact of Hurricane Harvey in Texas at the census block level. By integrating a bilevel learning framework within a deep neural network model, we specifically target and reduce spatial biases. Our results demonstrate that this framework not only improves prediction accuracy, particularly in areas with low population density, but also ensures more disaster response strategies with improved fairness. This study provides a novel contribution to the field by showcasing how AI models can be adapted to ensure fairness in disaster management, offering valuable insights for future AI-driven disaster assessment and response systems. Our results also promote the need for fairness in AI-driven disaster response to prevent unequal resource allocation, as evidenced in recent case studies on Hurricane Harvey
Data science · Econometrics · Economics · Estimation · Machine learning · Anomaly Detection Techniques and Applications · Computer Science · Data-Driven Disease Surveillance · Imbalanced Data Classification Techniques · Artificial Intelligence
The Data Revolution
Social Media in Disaster Risk Reduction and Crisis Management
Understanding communication dynamics on Twitter during natural disasters
European Union Regulations on Algorithmic Decision Making and a “Right to Explanation”
Who Tweets? Deriving the Demographic Characteristics of Age, Occupation and Social Class from Twitter User Meta-Data
Big Data's Disparate Impact
Mapping social vulnerability to enhance housing and neighborhood resilience
Communicating on Twitter during a disaster
Social media analytics for natural disaster management
Geographically Weighted Regression
Disaster resilience through big data
Investigation of social media representation bias in disasters
Spatial biases in crowdsourced data
Exploring flood mitigation governance by estimating first-floor elevation via deep learning and google street view in coastal Texas
Hazard risk awareness and disaster management
Rapid Damage Estimation of Texas Winter Storm Uri from Social Media Using Deep Neural Networks
Space, time, and situational awareness in natural hazards
The Digital Divide Among Twitter Users and Its Implications for Social Research
Geographic information science III
Enhancing population data granularity
Social media and disasters
Geographic Variation in Household Disaster Preparedness in the United States
Disaster Misinformation and Its Corrections on Social Media
Mining Twitter Data for Improved Understanding of Disaster Resilience
| Citation velocity | historical |
|---|---|
| Highly cited | No |