Sean M. Fitzhugh

Warning tweets: serial transmission of messages during the warning phase of a disaster event

Serial transmission – the passing on of information from one source to another – is a phenomenon of central interest in the study of informal communication in emergency settings. Microblogging services such as Twitter make it possible to study serial transmission on a large scale and to examine the factors that make retransmission of messages more or less likely. Here, we consider factors predicting serial transmission at the interface of formal and informal communication during disaster; specifically, we examine the retransmission by individuals of messages (tweets) issued by formal organizations on Twitter. Our central question is the following: How do message content, message style, and public attention to tweets relate to the behavioral activity of retransmitting (i.e. retweeting) a message in disaster? To answer this question, we collect all public tweets sent by a set of official government accounts during a 48-hour period of the Waldo Canyon wildfire. We manually code tweets for their thematic content and elements of message style. We then create predictive models to show how thematic content, message style, and changes in number of Followers affect retweeting behavior. From these predictive models, we identify the key elements that affect public retransmission of messages during the emergency phase of an unfolding disaster. Our findings suggest strategies for designing and disseminating messages through networked social media under periods of imminent threat.

Rumoring during extreme events: a case study of deepwater horizon 2010

Social scientists have proposed many different factors thought to influence rumoring behavior. Classical rumor theory points to the perceived importance, the level of uncertainty or ambiguity, and the potential to impact decision making as influential in determining the extent of rumoring. In this work, we test some of these proposed rumor determinants in the context of the the 2010 Deepwater Horizon oil spill, using data on communication dynamics from the popular microblogging service Twitter. Using a latent factor model, we measure rates of hazard-related conversation by exploiting joint variation in multiple conversation streams. Time series analysis of the resulting rates suggests that media coverage of the event is a major driver of rumoring behavior, supporting importance/saliency theories and disconfirming theories of information substitution for this event. Relevance of the event to decision making behavior also turns out to be an influential predictor in this case. Since information diffusion via serial transmission is a fundamental process by which rumors spread, we compare rates of serial transmission between control and hazard-related communication. Twitter posts are much more likely to be retweeted when they contain hazard-related keywords (versus control words). Implications of these findings for disaster response are discussed.

Tweeting the Spill: Online Informal Communications, Social Networks, and Conversational Microstructures during the Deepwater Horizon Oilspill

Informal online communication channels are being utilized for official communications in disaster contexts. Channels such as networked microblogging enable public officials to broadcast messages as well as engage in direct communication exchange with individuals. Here the authors investigate online information exchange behaviors of a set of state and federal organizations during the Deepwater Horizon 2010 oil spill disaster. Using data from the popular microblogging service, Twitter, they analyze the roles individual organizations play in the dissemination of information to the general public online, and the conversational aspects of official posts. The authors discuss characteristics and features of the following networks including actor centrality and differential mixing, as well as how structural features may affect information exchange in disasters. This research provides insight into the use of networked communications during an event of heightened public concern, describes implications of conversational features, and suggests directions for future research. DOI: 10.4018/jiscrm.2013010104 International Journal of Information Systems for Crisis Response and Management, 5(1), 58-76, January-March 2013 59 Copyright

A cross-hazard analysis of terse message retransmission on Twitter

Significance Online social networks (OSNs) enable time-resolved measurement of communication behavior during disasters, making it possible to probe the mechanisms by which messages are amplified or suppressed with precision unattainable by traditional data sources. To our knowledge, this research provides the first systematic study of the factors predicting the social amplification of risk communication in OSNs by examining the retransmission of official messages across five hazards. Our findings demonstrate the respective impacts of sender characteristics, message content, and message style in determining whether an official message will be passed on during an emergency, as well whether these vary across hazards. These results contribute to the evidence base for policies guiding the delivery by emergency management organizations of lifesaving information to the public. For decades, public warning messages have been relayed via broadcast information channels, including radio and television; more recently, risk communication channels have expanded to include social media sites, where messages can be easily amplified by user retransmission. This research examines the factors that predict the extent of retransmission for official hazard communications disseminated via Twitter. Using data from events involving five different hazards, we identity three types of attributes—local network properties, message content, and message style—that jointly amplify and/or attenuate the retransmission of official communications under imminent threat. We find that the use of an agreed-upon hashtag and the number of users following an official account positively influence message retransmission, as does message content describing hazard impacts or emphasizing cohesion among users. By contrast, messages directed at individuals, expressing gratitude, or including a URL were less widely disseminated than similar messages without these features. Our findings suggest that some measures commonly taken to convey additional information to the public (e.g., URL inclusion) may come at a cost in terms of message amplification; on the other hand, some types of content not traditionally emphasized in guidance on hazard communication may enhance retransmission rates.

Patterns of co-membership: Techniques for identifying subgraph composition

Abstract Different social processes give rise to network structures with distinctive properties. In this paper our goal is to identify the social processes that give rise to distinct network structures (specifically, subgroups). We examine particular structural meta-relations by identifying the properties of individuals associated with specific subgroups. Clues to the process of group formation and the context in which these groups form and persist may be extracted from the properties of individuals in those groups. Following this intuition, we propose a general technique for identifying systematic patterns of attribute occupancy to determine how individual attributes may drive group formation. To connect the social context in which groups form to their structural signatures, we relate subgroup composition to nodal attributes. We illustrate the utility of comparing subgroup (e.g., clique, n-clique, k-core, etc.) co-membership with nodal co-membership in a variety of attributes. The correlations between these two co-membership matrices illustrate clearly the strength of association between shared attributes and shared subgraph membership. Furthermore, examining these correlations across groups of different sizes indicates where these attributes are most strongly associated with group co-membership. Additionally, these correlations fit well into a QAP framework to determine where shared subgraph membership has a stronger (or weaker) relation to shared attribute membership than we would expect by chance. We demonstrate the technique with a series of large, online friendship networks on the order of thousands of nodes to illustrate how factors such as gender, cohort, residence, and other attributes are associated with co-membership across a range of clique sizes.