top of page

Amir Behzadan

Texas A&M University

Breakout Room 4.


Inclusiveness, Generalizability, AI/ML, Human-centered, Trust

How can the human dimensions of disaster impacts be more accurately captured and represented in the analysis, modeling and simulation of disasters?

I would like to approach this question from the perspective of data representation and inclusiveness. There are key challenges in accurately capturing and representing human behaviors, perceptions, and experiences in disaster events. First, there is the influence of the means of data collection on data quality and reproducibility. For example, there is evidence from the literature that suggests some disaster survivors (i.e., human subjects) prefer and experience lower distress levels during interviews and questionnaires compared to when psychological assessment measures are used. Additionally, in any participatory process that depends on an open invitation for participants, data describing human dimensions of disaster impact may be prone to self-selection bias, which occurs when the group that chooses to participate in the study is not equivalent (in terms of the research criteria) to the group that opts out. Some of these negative effects can be controlled prior to data collection using more rigorous experimental designs and constructs that consider variability in the population and propose alternative methods to reach out to disaster survivors, obtain informed consent, and ensure participant retention. Along these lines, some studies have suggested that instead of compensating human subjects, researchers could connect them with information or refer them to proper resources, thus promoting equity in access particularly among underserved communities that are sparsely represented in study samples. Post-data collection, the internal and external validity of human data must be carefully examined using means such as bootstrapping, while also controlling for moderating factors and correlations among variables of interest.

What type of data and supporting research infrastructure would be necessary to enable novel, transdisciplinary approaches to answering these and other human-centered disaster questions?

The impact of disasters on humans often comes across as an “after-thought”, thus limiting our ability to create human-centered approaches to disaster management and mitigation. As an educator with background in civil engineering who is heavily involved in disaster research, I often find a large portion of the disaster community lacking true transdisciplinary perspectives and motivation to tackle grand challenges in this domain. The most important research infrastructure that is needed in this regard is a common language/ontology that draws from multiple disciplines to clearly define key terms and concepts, which hinders clear communication and transfer of knowledge among disciplines. To date, there is still debate about the true definition of basic terms such as resilience, stakeholder, victim vs. survivor, underserved vs. underrepresented, disaster vs. hazard, and many more. Secondly, there is a lack of a common data sharing and analytics platform in disaster events when multiple stakeholders and data modalities are involved. Disaster data comprises both qualitative (often collected through direct interaction with humans) and quantitative data (often captured by sensors and technologies embedded in the built environment). The variability in the spatiotemporal resolution of multiple datasets, inconsistency in data curation and labeling methods (e.g., for training AL/ML models), along with the ephemeral nature of disaster data makes it challenging to reconcile disparities, analyze such data in a single, unified framework, and ultimately produce reliable and actionable information for decision-making (previous studies on disaster cleanup, for example, have reported 50% error in estimated debris volumes after flood events).

In what ways can US-Japan collaborations advance these questions in new and important ways?

Both Japan and the U.S. are technologically advanced societies (as a whole) which explains the existence of a relatively large body of literature in each country with respect to disaster mitigation technologies (e.g., satellites, drones, robots) and infrastructure (e.g., warning systems). However, in the context of human-centered approaches to disaster resilience, it would be fascinating to study the role of cultural and societal factors between Japan and the U.S., and how they might impact the successful design, implementation, and generalizability of various disaster management practices within and across the two countries. When it comes to human data collection, for example, researchers need to consider the fact that some populations (i.e., Japanese) feature more subtle communication versus others (i.e., Americans) that tend to be blunter. Likewise, human-centered models must be aware of the more distinctive role that gender, age, and social hierarchy play among Japanese communities, while many disaster-prone communities in the U.S. are home to socioeconomically diverse populations and first-generation immigrants. From a team dynamics perspective, there are several interesting questions that can be investigated considering that Japan's culture is primarily collectivist (potentially impacting their perceptions on issues such as vulnerability and fair allocation of disaster resources), compared to the individualistic culture that is predominant in the U.S. Trust in technology and authority is another avenue for cross-sectional and longitudinal research in Japan and the U.S. These potential research directions can bring together social scientists, political scientists, economists, and engineers to identify and work on new opportunities and U.S.-Japanese partnerships.

bottom of page