Expressions of risk awareness and concern through Twitter: On the utility of using the medium as an indication of audience needs
Introduction
Social networking sites are increasingly gaining credibility as information sources. Microblogging sites such as Twitter allow us to access information from both official organizations and public citizens, instantaneously covering events as they happen. At the same time, the information broadcast on Twitter may help to provide a sense of the needs and responses of those affected by large scale events as they are taking place. Thus, the possibility exists that crisis managers and emergency responders may be able to use Twitter as a bellwether for the specific needs and concerns aired by varying publics during crises and disasters.
The current study offers a content analysis of information that was tweeted in the time leading up to the landfall of Hurricane Sandy in 2012. It describes the relative frequency of expressions of risk awareness and concern, building off of past psychometric work in risk analysis categorizing different types of risk perception. It explores differences in these risk perceptions across age, sex, ethnicity, and language group. These differences are discussed in the utility of the study of Twitter content in evaluating the needs and concerns of different sectors of the audience, and the ways in which this information may be used to inform emergency management operations.
A substantive body of work in the risk communication literature has examined risk perceptions primarily in terms of fear (Reynolds & Seeger, 1995). This is somewhat problematic, as this literature often works under the assumption that the induction of any negative emotion can be qualified as fear, and that any type of fear will have relatively similar effects on response to one’s surroundings. (Dillard et al., 1996, Higbee, 1969).
More recent research on audience responses to threats and risks has moved toward a more sophisticated understanding of the concept (Trumbo & McComas, 2003). For example, one research program has argued that responses to risk can be broadly categorized into two dimensions: knowledge and dread (Slovic, 1994). Dread is defined as the perception of the likelihood that a risk will harm those who are not directly involved. Conversely, knowledge is the cognitive understanding of the effects that a risk may have on those whom it will affect. Trumbo and McComas (2003) argue for another dimension of risk perception, this one detailing the perception of the number of people who are likely to be affected by the risk at hand; this is consistent with other research in the risk analysis canon (Slovic, Fischhoff, & Lichtenstein, 1986).
Sandman’s (1999) Risk = Hazard + Outrage offers similar arguments for the multi-faceted nature of risk perception, arguing that alarm, magnitude, and probability are critical components in what we can expect from audiences under duress. Sandman, Miller, Johnston, and Weinstein (1993), add that outrage, or the affective experience of risk perception, may include a range of emotions such as anger, suspicion, fear, distrust, and contempt.
Evaluating these affective responses associated with risk perception is critical in determining how to respond, as they are likely to moderate cognitive perceptions of the risk at hand and subsequent responses to that risk. Some research indicates that those experiencing negative affect of one kind or another may be likely to overestimate the potential impact of a threat, particularly if that threat involves a loss of personal control (Johnson and Tversky, 1983, Lerner and Keltner, 2000).
Complicating matters further is a long history of research indicating differences in disaster preparation, information needs, and likely responses across demographic and cultural groups. These differences in response are likely to beget different psychometric responses, which are key considerations for emergency planners and responders attempting to motivate people to action. In terms of preparedness, Caucasians tend to be better prepared for crisis and disasters than members of other communities, likely due to economic advantages and a history of structural inequality among non-whites (Fothergill et al., 1999, Spence et al., 2008).
Past scholarship on ethnicity and crisis response also suggests that inequities are critical in terms of evacuation and access to basic relief supplies (Gladwin and Peacock, 1997, Peacock, 2003). There is also evidence of difference across ethnicity in risk perception and response. Hohm (1976) offers data suggesting that during crises African Americans tend to indicate less concern than do Caucasians. Other studies suggest similar discrepancies in risk expression regarding environmental risks, and argue this lack of concern is related to a lack of knowledge or understanding about the risk at hand (Van Ardosol, Sabagh, & Alexander, 1965).
Cultural differences may also be evident in the acceptance and belief in risk messages, and this may translate into very different reactions in the face of an impending crisis. Evidence suggests that members of historically underrepresented groups may be less likely to accept risk messages as credible, or to act upon those messages, and that they may not do either until verifying the information through trusted interpersonal channels (Spence, Lachlan, & Griffin, 2007). This need for interpersonal confirmation leads to delays in response time (Fothergill et al., 1999, Lindell and Perry, 2004).
Further, there may be specific underrepresented communities who are less likely to consume English language media, which leads to challenges in the dissemination of risk messages. During Katrina, nearly a third of the Latino population of New Orleans was identified as speaking English “less than very well” (Hilfinger-Messias & Lacy, 2007). Latino respondents affected by Katrina were likely to indicate that they were not adequately prepared and/or were unaware of the full implications of not evacuating (Hilfinger-Messias & Lacy, 2007). Another study suggests that about 32% of the population affected by Hurricane Andrew was heavily dependent on Spanish language media for information concerning the storm (Gladwin & Peacock, 1997).
A significant body of research has also suggested gender differences in the information needs and concerns aired during and after crises and emergencies (Seeger, Venette, Ulmer & Sellnow, 2002). For example, one study examining information seeking following a terrorist event (Keinan, Sadeh, & Rosen, 2003) suggested that men may gravitate toward visual media, while women preferred print and radio. Research also suggests that women are more inclined to seek out crisis information from social networks (Turner, Nigg, Paz, & Young, 1981), and then turn to media for confirmation. It seems plausible then that Twitter, as a medium, could be especially attractive to women due to its social networking characteristics.
There are also specific, predictable differences between men and women in terms of their informational needs following disasters and emergencies. For example, after Katrina women reported that they were more inclined than men to seek information concerning evacuation attempts, rescue efforts, obtaining shelter, the location of friends and family, the impact of the storm on others, the widespread impact of the storm, and where to obtain medical care (Spence et al., 2008).
Further, research suggests that men may be more likely to attend to crisis information involving behavioral advocacy, or those that will lead to specific actions (Hoffner, Fujioka, Ibrahim, & Ye, 2002). At the same time, however, some research indicates that males are more likely to channel information concerning a crisis or emergency event into blame or outward expressions of hostility or aggression toward others (Baukus and Strohm, 2002, Brody and Hall, 2000, Greenberg et al., 2002). In short, women seem more inclined to seek out information pertaining to survival and emotional needs, while men seem more inclined to channel any information received toward anger or hostility.
There is a significant threat posed by the failure to consider the informational needs and desires of diverse audiences during crises and disasters. Singularized messages that fail to address specified needs may lead to scenarios in which some affected audiences are ill informed, confused, or resort to antisocial responses, thus actually increasing the level of harm presented by the crisis at hand. Sandman (2003) and colleagues have argued that crisis messages should alert and reassure people at the same time, and that they should induce an appropriate amount of fear among those affected – enough to adequately motivate the audience to action, but not so much that they resort to antisocial behavior or abandon hope altogether. The goals are expressed through the conceptual formula Risk = Hazard + Outrage. This program of research defines Hazard as the assessment of risk parameters associated with the crisis, while Outrage is the degree of negative affect that may be associated with responses to the event (Lachlan and Spence, 2007, Sandman et al., 1998, Sandman et al., 1987). Others have extended the definition of Outrage to include the specific actions of individuals to the messages they receive (Lachlan and Spence, 2007, Lachlan and Spence, 2010). If audiences are not responding appropriately, and this can be attributed to a lack of knowledge about the parameters of the event, then they should be provided with information. Once that vector is addressed, then negative affective responses can be considered.
Within this framework, specific behavioral advice couched in a degree of alarm that is enough to motivate but not to paralyze should lead audiences to make ideal decisions concerning the protection of life, health, and property. Of course, what constitutes an appropriate level of negative affect is likely to vary across different groups, as are the specific types of information that may be necessary or desired. Risk messages that encourage tangible responses, alongside a moderate level of negative affect, will engender empowerment, and individuals will be more inclined to act the manner desired due to their perception of control over their own fate. (Seeger, Vennette, Ulmer, & Sellnow, 2002). Failing to do so may be disastrous. Further complicating the matter is the placement of the appropriate information for a specific audience within a morass of Tweets related more to affective release.
Past research supports the notion that those affected by crises will use Twitter as a means of affective release, and that these expressions of affect may be indicative of the collective concerns of one or more groups of people. Under duress, individuals are likely to use social media to release emotion and to connect with others. The more affective functions of social media – both in terms of the request for and granting of social support – are likely to nurture involvement with others. For example, one study reports that during the Arab Spring uprising, affective displays became increasingly more intense as protests escalated toward political upheaval (Papacharissi & de Fatima Oliveira, 2012). In another, as a natural disaster moved toward its climax, those affected were less likely to use Twitter to offer information related to mitigation, and more likely to use the medium as an outlet for expressing the tremendous levels of anxiety and fear they were experiencing (Spence, Lachlan, Lin, & Del Greco, in press). Some have also argued that Twitter users have a higher tendency than most to display narcissistic tendencies (McKinney, Kelly, & Duran, 2012). These narcissistic tendencies may motivate them toward sharing emotions and details of their experiences, in the hopes of receiving some type of validation from others.
Regardless of the source of motivation, it is clear that Twitter users are likely to offer specific indications of their risk perceptions, particularly those that are related to negative affect. As detailed elsewhere in the risk communication literature, and understanding of the affective needs of affected populations and subpopulations may be critical in planning for and responding to emergencies.
Clearly, differences in informational needs and affective responses can be expected across audiences from different cultural backgrounds, socioeconomic levels, and across gender. Individuals will likely experience different needs and offer different responses given the particular constellation of variables that can be attributed to them. At the same time, risks can manifest in different psychometric properties, and these properties may be a product of these specialized needs.
For a number of years, researchers, government agencies, and emergency management organizations have been attempting to identify the needs of varying subpopulations, and identify best practices for doing so. As the public becomes more and more dependent upon social media for satisfying information needs under a variety of circumstances, it becomes apparent that social media content may serve as a good barometer for evaluating the needs of those affected.
To that end, the current study builds upon the work of Spence et al. (in press) by offering an initial exploration of the utility of Twitter content in determining the needs and responses of diverse respondents under the duress of an impending crisis. In the days leading up to the landfall of Hurricane Sandy, large quantities of tweets were examined at varying time points, in order to get a sense of the specific risk perceptions offered by varying groups. Given what is known from previous research concerning potential differences in risk perception across demographic, the following research questions are proposed: RQ 1 Will the specific types of affective responses reported in tweets vary between the sexes? RQ 2 Will the types of affective responses indicated vary across age range? RQ 3 Will the types of affective responses indicated vary across age ethnicity? RQ 4 Will the types of affective responses indicated vary across the language in which the tweet was sent?
Section snippets
Data collection
Hurricane Sandy served as the backdrop to the current study. Tweets were harvested using TweetArchivist (www.tweetarchivist.com), a software program that allows users to compile recent tweets using numerous search criteria. In the four days leading up to landfall, Tweets were identified and compiled that contained the hashtag sandy. This hashtag was selected because it was used by multiple emergency relief organizations at the federal level to identify messages related to the storm.
Analysis plan
For all four research questions, a series of chi-squared analyses were conducted to explore relationships in the data concerning the specific types of affect suggested by the tweet content. Because fear, magnitude, dread, anger, risk, fear of losing assets, and fear for health were all coded as separate binary variables, for each research question a separate chi squared analysis was conducted for each dependent variable. Thus, seven separate analyses were conducted for each research question.
Discussion
The current study aims to explore differences in risk perceptions across demographic attributes in the days leading up to the landfall of hurricane Sandy. More broadly, the study aimed to determine whether Twitter can be used by researchers and emergency responders as a means of detecting differences in risk perceptions across diverse audiences. The results indicate some similarities and differences in the perceptions of risk, when compared to past literature in crisis and risk communication.
Conclusion
Although only an initial exploration into the utility of Twitter as a tool for informing crisis communication messages, the results offer some initial insights into its potential. For the most part, the findings are consistent with past research suggesting structural differences across audiences and subpopulations in terms of their risk perceptions and responses. At the same time, there are a handful of findings that deviate from previous research, and provide surprising insights into audience
References (40)
- et al.
Gender differences in perceptions of media reports of the Gulf and Afghan conflicts
- et al.
Gender, emotion, and expression
- et al.
The multiple affective outcomes of AIDS PSAs: Fear appeals do more than scare people
Communication Research
(1996) - et al.
Race, ethnicity and disasters in the United States: A review of the literature
Disasters
(1999) - Freelon, D. G. (2008). ReCal2: Reliability for 2 coders. Online kappa calculator. Retrieved...
- et al.
Warning and evacuation: A night for hard houses
- et al.
Diffusion, media use and interpersonal communication behaviors
Fifteen years of fear arousal: Research on threat appeals: 1953–1968
Psychological Bulletin
(1969)- et al.
Katrina-Related health concerns of Latino survivors and evacuees
Journal of Health Care for Poor and Underserved
(2007) - et al.
Emotion and coping with terror
A human-ecological approach to the reality and perception of air pollution, The Los Angeles case
Pacific Sociological Review
Affect, generalization, and the perception of risk
Journal of Personality and Social Psychology
Attitudes and reactions to media coverage of terrorist acts
Journal of Community Psychology
Content analysis: An introduction to its methodology
Hazard and Outrage: Developing a psychometric instrument in the aftermath of Katrina
Journal of Applied Communication Research
Communicating risks: Examining Hazard and Outrage in multiple contexts
Risk Analysis
Sex differences in negative psychological responses to crisis news: The case of the I-35W collapse
Communication Research Reports
Beyond valence: Toward a model of emotion-specific influences on judgment and choice
Cognition and Emotion
Communicating environmental risk in multiethnic communities
Narcissism or openness?: College students’ use of Facebook and twitter
Communication Research Reports
Cited by (98)
Trends in bushfire related tweets during the Australian ‘Black Summer’ of 2019/20
2023, Forest Ecology and ManagementData mining of social media for urban resilience study: A case of rainstorm in Xi'an
2023, International Journal of Disaster Risk ReductionCombining multiple data sources to identify actor involvement in environmental governance: Wildfire in the American West
2023, Environmental Science and PolicyImproving search for gasoline during a hurricane evacuation event using social media
2023, EURO Journal on Transportation and LogisticsOpening organizational learning in crisis management: On the affordances of social media
2021, Journal of Strategic Information Systems