Search for: All records

The widespread use of Artificial Intelligence (AI) has highlighted the importance of understanding AI model behavior. This understanding is crucial for practical decision-making, assessing model reliability, and ensuring trustworthiness. Interpreting time series forecasting models faces unique challenges compared to image and text data. These challenges arise from the temporal dependencies between time steps and the evolving importance of input features over time. My thesis focuses on addressing these challenges by aiming for more precise explanations of feature interactions, uncovering spatiotemporal patterns, and demonstrating the practical applicability of these interpretability techniques using real-world datasets and state-of-the-art deep learning models. 

The concept of resilience is surging in popularity, but relevant discussions are often disconnected from one field to another. To prompt integration of disparate conversations on resilience, we examine the concept’s origins etymologically, genealogically, and by analyzing the interdependencies of drinking water and public health systems in six academic disciplines and practice-oriented fields. These disciplines are engineering, social work, urban studies, political science, communication, and public health. While the disciplinary resilience literatures are relatively stove-piped from one another’s contexts, they all theorize resilience at multiple levels of analysis. They also engage a range of understandings of how to build resilience in complex systems. This paper brings several conversations together, addressing gaps and resonances in disciplinary conceptualizations of resilience with nine propositions to cultivate interdisciplinary and transdisciplinary discussions and debates. We ground this creative inquiry in real-world examples of water system crises to highlight subthemes among the propositions and stimulate more diverse discussions moving forward. We examine dynamics of interfaces and interactions within and between systems through the Elk River Water chemical contamination in Charleston, West Virginia in 2014. We investigate tensions that arise in knowledge and practice through lead poisoning of public water systems in Washington, D.C. and Flint, MI. Finally, we consider how change and persistence shape learning through water infrastructure in Southern California. All together, these propositions offer a starting point and a provocation to strengthen theorizing around resilience for critical infrastructure systems. 

The Safe Drinking Water Act Public Notification Rule requires that customers of public water systems (PWS) be informed of problems that may pose a risk to public health. Boil water advisories (BWA) are a form of communication intended to mitigate potential health risks. The Centers for Disease Control and Prevention (CDC) developed guidance for BWAs. We examined how local US news media incorporate the CDC’s guidelines when reporting on BWAs. A content analysis of 1040 local news media articles shows these reports did not consistently incorporate CDC guidelines. Overall, 89% of the articles communicated enough information for readers to determine if they were included in the impacted area. Articles that included at least some of the CDC’s instructions for boiling water were likely (p < .001) to include other risk information, such as the functions for which water should be boiled (e.g., drinking, brushing teeth) and that bottled water could be used as an alternative source. However, this information was included in only 47% of the articles evaluated. Results suggest public notifications often do not serve the public need for clear risk communication. 

Abstract This research investigates medium‐scale disruptive events to understand how these events influence communication and coordination between two interdependent systems (i.e., the water system and the public health system). Medium‐scale events are events that are often overlooked as routine as they occur with more frequency than large‐scale events, yet they have the potential to provide important information about the state and vulnerability of systems, and, if not managed appropriately, can cascade into larger‐scale crises. A survey of US public drinking water systems (N = 471) shows that medium‐scale events promote coordination, especially when those events have a public dimension. Findings also reveal that several features of water systems including surface water sources, system size, and ownership types are associated with higher levels of interaction with the public health systems. Additionally, a network analysis identifies three distinct subnetworks that engage in emergency response activities. The strength of the working relationship was strongly associated with coordinated emergency responses, coordinated public responses, planning, and technical assistance. Findings have implications for both theory and crisis management.