A framework for understanding online group behaviors during a catastrophic event

Highlights

• We proposed a novel framework to analyze online group behaviors and the emergent communication networks in an emergency.

• Group interactions between media, private organizations, emergency agencies, and the public were analyzed.

• Each group during hurricane Harvey has its unique structural and semantic interaction patterns.

• More frequent group interactions between the public and private organizations are observed.

• The interactions between the public and private organizations have the highest topical usefulness and topical difference.

Abstract

This study investigated the underlying mechanisms of online social media group behaviors in an emergency. The proposed framework was designed to analyze group behaviors/interactions and examine the main topics of interest among numerous tweets generated in an emergency. We collected tweets sent during Hurricane Harvey in 2017 and applied the framework to demonstrate its effectiveness. The proposed framework enables us to understand the unique characteristics of group interactions and develop operational strategies to effectively communicate with the public, as well as other groups, as critical emergency information appears in an online social network.

Introduction

Communication during and immediately after a catastrophic event is a core component of effective emergency response and recovery (Houston et al., 2015; Nowell & Steelman, 2015a; Scholl & Patin, 2014). When a catastrophic event causes threats to human life and property, individuals need information about what has happened and what is still ongoing within a disaster-affected area and beyond (Rodríguez, Diaz, Santos, & Aguirre, 2007). Organizations and the public work together to fill information gaps through creating and sharing emergency information (Liu, Fraustino, & Jin, 2016). Such communication connects affected people, families, and communities with first responders, support systems, and other family members.

In terms of concept, purpose and usage, structure, formation, functionality and interactivity, traditional communication channels have been supplemented by emerging online communication channels such as Facebook, Twitter, and YouTube. The two main advantages of the new channels are that an information generator can directly and rapidly deliver information to end-users and facilitate two-way communication between information generators and receivers. The interactions via social media during an emergency create an enormous dataset and network (Alam, Ofli, & Imran, 2018; Imran, Mitra, & Castillo, 2016; Nguyen, Ofli, Imran, & Mitra, 2017). The use of social media also enables scholars to investigate disaster impact analysis, social behaviors and patterns using the dataset generated during an emergency. Twitter, in particular, has been popular for disaster research, due to the large number of users and the network’s application programming interface (API) (Makice, 2009; Twitter, 2019b). Twitter is a micro-blog social media service where users read and write short messages. Twitter is available via its website, short message services, and mobile applications. As of the 1 st quarter of 2017, Twitter had more than 330 million monthly active users, with more than 1 billion monthly visits to the website. According to the report by InternetLiveStats (2017), Twitter publishes around 200 billion tweets each year, or approximately 6500 tweets per second. As of December 2016, Twitter had more than 67 million active users in the United States alone. Twitter also launched Twitter API, which enabled users to collect tweets via keywords or locations.

Networks like Twitter help to predict and understand disasters such as hurricanes and earthquakes. For example, a variety of studies (Robinson, Power, & Cameron, 2013; Toriumi et al., 2013; Umihara & Nishikitani, 2013; Yates & Paquette, 2011a) have investigated the use of social media information sharing patterns and the ways in which social media was used for decision-making at critical junctures during disasters. The contents of such tweets have also facilitated understanding of disasters and corresponded to the existing theories using natural language processing techniques. One of the research findings is that keywords in emergency-related tweets showed higher correlations during disaster events (Earle, Bowden, & Guy, 2011). Guan and Chen (2014) also demonstrated that there was a close connection between the activities on social media and the extent of disruptions related to Hurricane Sandy.

Social media play a critical role in disaster management by facilitating communication between governmental agencies, organizations, and the public during a catastrophic event (Kryvasheyeu et al., 2016; Lindsay, 2011a). The use of social media in an emergency could be conceptualized into two broad categories: 1) sharing emergency-related information and 2) systematic practices (Lindsay, 2011b). Emergency-response agencies use social media to share information and receive user feedback/requests via messages and wall posts (Feldman et al., 2016; Kim & Hastak, 2018c, 2018b). For example, the U.S. Federal Emergency Management Agency (FEMA) operates the social hub page to disseminate weather and disaster information and share emergency information from state and local emergency agencies with the public (FEMA, 2018a). With regard to systematic practices, several studies have explored the effective systematic use of social media and data by applying multiple methods, including network analysis, semantic analysis, and content analysis (de Albuquerque, Herfort, Brenning, & Zipf, 2015; Gao, Barbier, & Goolsby, 2011; Imran, Elbassuoni, Castillo, Diaz, & Meier, 2013; Scott & Errett, 2017; Yin et al., 2015). Social media data have been used to develop early emergency warning systems, situation awareness systems, and citizen communication platforms (Imran, Castillo, Lucas, Meier, & Vieweg, 2014; Kryvasheyeu et al., 2016; Lu et al., 2015).

Recently, FEMA recommended that state and local emergency agencies open social media accounts. They also opened social media in emergency management (IS-42) under the independent study program (FEMA, 2018b). For example, Texas is the second-most populous U.S. state. According to the U.S. Census Bureau, the population of Texas is 28,304,596 (U.S. Census Bureau, 2018). Of the total population, 61.3 % (17.3 million) are over 18 and below 65 years old. As shown in Table 1, the number of followers/subscribers of @FEMAHarvey and @FEMARegion6 were very few: 10,678 and 20,100, respectively. That is 0.0006 % and 0.0011 % of the adult population of 17.3 million. There may be several reasons for these low numbers. The public might already be linked with different groups, such as media and organizations, which provide reliable information, or they can access emergency information from offline communication channels. The outcomes of this research could enable emergency agencies to develop social media operation policies and strategies for effective disaster response programs.

To utilize online social media as a core of emergency communication channel, it is critical to identify the characteristics of both online user and group behaviors and their communication patterns during a catastrophic event. It enables emergency agencies and related organizations to develop effective social media strategies and operations.

Therefore, the objective of this study is to develop a framework to understand online user and group behaviors and characteristics during a catastrophic event. It includes multiple methods to analyze emergent user networks, behaviors and messages within/across different groups: the public, private organizations, emergency agencies, and media. We collected Twitter data during Hurricane Harvey and applied the proposed framework. The remaining sections of this manuscript are composed of following sections: Literature review (Section 2), theoretical basis (Section 3), and methodologies (Section 4). Then, results and discussions (Section 5) and implications and limitations (Section 6), and conclusions (Section 7) will be described.