Natural disasters and Twitter: Thinking from both sides of the tweet
First Monday

Natural disasters and Twitter: Thinking from both sides of the tweet by Julia Skinner



Abstract
This paper is a reflection on the process of engaging with Twitter from the perspective of a researcher and a user. It includes the author’s experiences as a content producer on Twitter during wildfires in Colorado, followed by the experience of researching tweets produced during Hurricane Sandy. The goal is to outline the author’s subjective understanding of both experiences, and how those experiences play off each other to inform future social media use and future research.

Contents

Introduction
The user experience
The research experience
Thinking from both sides: How does research affect online behavior?
Thinking from both sides: How does this approach affect future research?
Conclusion

 


 

Introduction

Twitter has become an important tool for disseminating information during natural disasters, because of the real–time nature of updates and the fact that updates are publicly available. Numerous studies have been conducted from a variety of perspectives in order to understand Twitter’s role in natural disaster–related communication. Some have argued that it is an opportunity for citizens to engage as content creators during news events rather than simply serving as consumers (Freeman, 2011). One study (Gao, et al., 2011) found that Twitter was also being used as a way to determine what resources were needed in disaster areas. Others have looked at social media use by first responders (Latonero and Shklovski, 2011), and Mills, et al. (2009) argued it is not yet being used in a way that is reliable or effective, although it has the potential to be useful. Condry (2011) combined Twitter with other social phenomena, arguing that music and social media could be the basis for cultural movements (e.g., political protests or other large–scale events) during disasters. Utz, et al. (2013) contend that the type of social media platform used (Twitter, Facebook, etc.) has a stronger impact on crisis communication than the type of crisis taking place. While this work helps inform our understanding of social media use and disasters, it is missing a subjective perspective on the experience of communication via Twitter during such an event. It also does not yet include studies related to very recent phenomena, such as Hurricane Sandy.

Even with the variety of studies examining social media and natural disasters [1], none of these directly reflect upon the authors’ experiences as consumers and creators of that media. That kind of self–reflection provides valuable subjective, qualitative data that can help us to understand not only what is being shared within social media spaces, but why those sharing behaviors might exist. It also helps the researcher critically examine the experiences they bring to their work, and reflect on how their work shapes their media use, and vice versa. This paper takes that approach by reflecting on the author’s time as a Twitter user engaged in disaster tweeting (defined here as the dissemination of relevant information via Twitter during a natural disaster), and as a researcher studying an entire body of disaster tweets.

This methodology, in which social media use and social media research engage in an iterative process and inform each other, has a variety of other implications for future research. According to Ahn (2013), data analysis and social media both inform each other: Data analysis helps us understand how people participate in these spaces and how their engagement is increased, while social media can be used as a knowledge source in which researchers and users both draw from social wisdom to improve services via crowdsourcing. Studies that bring forth new ways of examining this engagement are particularly valuable, as Twitter communication patterns are global in nature (Leetaru, et al., 2013), meaning that natural disaster information can come from users far away from the disaster, which could potentially lead to different patterns of engagement and a wider variety of perspectives being shared.

The value of subjective experience as a tool to provide depth to complement the breadth of quantitative approaches has been outlined in numerous studies on qualitative research methods [2]. While much research focuses on the lived experience of the research subject as a vantage point for observing a phenomenon [3], there is literature pointing to the importance of the researcher’s experience as a vantage point as well as a potential change agent within the area of interest.The inclusion of authors’ personal experience has been advocated for by Krefting (1991), who synthesized earlier work and found that qualitative research can be strengthened through reflexive practices in which the researchers are transparent about the effect of their own experiences on the work in question, and can learn more about themselves and their impact on their studies by continuously journaling their experiences during the study process. Additionally, Brettell’s (1993) edited volume on ethnographers’ experiences with research subjects reading study results or otherwise learning what was said about them outlines the researcher’s position as a participant in the work, rather than simply an observer. This collection of perspectives highlights the role of the researcher as being connected to the subjects. It also stresses that the researcher’s self–reflection, and reflection on the contexts in which researchers operate during a study (including what to include and exclude in publications), can have an impact not only on subjects’ perception of the researcher and their work, but on their understanding of the contexts in which they intersected with that researcher (e.g., community, workplace, school). This paper brings together the lived experience of the user with that of the researcher in order to provide a subjective perspective that synthesizes multiple viewpoints.

 

++++++++++

The user experience

When there are fires in my home state, one way I try to help is by using my research skills to find timely information and share it with people who need it. By far the most effective way I’ve found to do this is via Twitter, using whatever hashtag is being used for the fire (users in Colorado seem very consistent with this — there are rarely more than one or two hashtags associated with a single fire). I share Office of Emergency Management (OEM) updates, Google Crisis maps, updates from local animal shelters, news stories, and other content directly related to that disaster, all with the end goal of keeping people within the impacted area informed. At first I felt self–conscious about tweeting about a natural disaster that was far away, but it was the only way I could think to help, and feedback from fellow users has indicated its effectiveness. For example, a user might thank me for sharing a resource in a reply, or might modify my tweet in a way that indicates the value they place on the resource (e.g., “Good to know MT @BookishJulia: new boundaries for #fourmilefire. Check the crisis map!”) [4] The importance of using Twitter in this way was reinforced last summer when I received a direct message thanking me for tweeting the latest boundaries of a fire evacuation zone. The sender’s house was in the area, but she had not gotten the call to evacuate. Because she checked Twitter and saw her neighborhood was evacuating, she had a chance to go back and save her pets and some belongings before the roads were closed.

I have found that my tweets sharing practical resources (Google crisis maps, OEM updates, updates from animal shelters, etc.) tend to be the ones that are shared most often. For example, the tweet “Folks near #flagstafffire: here’s how to get calls from the emergency alert system (esp. for those w/ cell phones) bit.ly/MAPRBd” [5] was retweeted four times, while a question about whether a Google Crisis Map had been created for the Waldo Canyon Fire (under #waldocanyonfire on Twitter) received very little attention and no retweets [6].

While practical information tends to be most widely shared, I am expected to hold to the same naming conventions as other Twitter users engaged in disaster tweeting related to fires. For example, while the above status was retweeted four times, a very similar status [7] from the day before was not, because I labeled it as “#flagstaff fire” instead of the accepted hashtag (#flagstafffire). Twitter users sharing information on Colorado wildfires tend to adopt a hashtag early, use it consistently, and encourage others to do the same (e.g., tweeting “Remember, the hashtag we’re using is #fourmilefire: Make sure to include it when you have something to share!”) This means a large amount of information about that disaster is grouped together. In my early experience, it also meant that when I did not use the same hashtag as other users my tweets were less visible because there were not multiple terms being searched.

During and after each disaster, I also gain followers who live and work in disaster areas, including some emergency responders. At this point, after tweeting information about fires near Boulder, Denver, and Colorado Springs, I’ve noticed enough interest in my tweeting that it has served as an impetus to continue using Twitter as a space to share information about natural disasters. The number of retweets of my posts are still minimal (most are shared two–five times, with the more popular ones being shared around 10 times), although as my network of connections increases there are more modified tweets of my posts, which shorten a tweet to include other information.

When Hurricane Sandy began moving toward New York, I chose to share information about that disaster as well. Hurricane Sandy was interesting because, unlike my tweets about wildfires, many of the resources I tweeted for this disaster were not shared as widely (most were not retweeted at all). Even in my first disaster tweeting experiences many of the resources I shared were retweeted, which led me to believe that the context of Sandy tweets (which focused on a much larger and more heavily populated area) made it easier for my tweets to be buried in a rapidly moving information stream. I had become so accustomed to receiving a positive response to my tweets that I found myself tweeting less often about Sandy as time passed, after a lack of feedback made me unsure whether or not what I was sharing was helpful. Our research on Hurricane Sandy offered me a chance to reflect on this process, and to contextualize my individual activities within the scope of Sandy tweets as a whole.

 

++++++++++

The research experience

Because of the recent experience of Hurricane Sandy, research on the storm has yet to appear in scholarly journals. However, the Pew Research Center’s Project for Excellence in Journalism (2012) released their findings on Twitter use during the storm. Between 29 October and 31 October 2012, the tweets they examined included a variety of content, including news and information, photos, videos, and jokes. Twitter experienced a surge of use during this time, which the researchers suggested was due to the loss of power by individuals and news agencies, who then relied upon the service to share and receive information.

Our research divided the tweets from Hurricane Sandy by the four phases of emergency management (FEMA, 2011). We sampled from each phase and conducted qualitative content analysis to determine what information was being shared, and if there were differences in the content of tweets between the different phases. Because this is an in–progress study done with a research team, the description of overall findings in the present paper will be minimal. Instead, the focus here is on my experience with the research and how the findings relate to my personal Twitter use.

The goal of the study was to understand how Twitter was used to share information about Hurricane Sandy during the different stages of that disaster. Because of my experiences, I approached the research project with a perspective similar to Gao, et al. (2011), and expected our samples to be dominated by tweets containing relevant and timely information, as well as requests for assistance. However, we found that overall, these types of tweets do not make up the majority of original posts or interactions.

Similar to the Pew (2012) study, we uncovered wide variety in tweet content, and we found that this content varied over the course of the disaster. As Hjorth and Kim (2011) found, we saw Twitter as a place for sharing affective information. However, Hjorth and Kim found that it was a place to express grief after a disaster, while our research was primarily concerned with the time during the disaster, meaning that the type of affective information was different. In our case, it consisted of well wishes for those in disaster areas, and humorous posts about the hurricane.

Utz, et al. (2013) found that crisis communication via social media was more effective than other types of media. However, people still consider traditional media (e.g., newspapers) as more credible. This is important for us as researchers, because it shows that this emerging area has great promise but is still not fully adopted or trusted. If we think about this in terms of diffusion of innovations (Rogers, 1995), this evidence points to Twitter as being in the middle of the innovation bell curve. Real–time updates on Twitter during natural disasters, political unrest, or any other time–sensitive event show the value of this tool for disseminating needed information, and may help with more widespread adoption.

 

++++++++++

Thinking from both sides: How does research affect online behavior?

Seeing the wide variety of Sandy–related content being posted to Twitter gave me the opportunity to appraise my own Twitter behavior, and think about how I can use Twitter as a tool to disseminate disaster information more effectively. Several trends emerged from our data that may be helpful for developing Twitter as a disaster preparedness tool.

The most obvious trend is the sheer volume of information that flows through Twitter: In 2011, users sent an average of 140 million tweets per week (Twitter, 2011). With that high volume of information, it becomes easy for individual tweets to be lost in the stream. Our study and Pew’s (2012) research both uncovered considerable variety in the content of Hurricane Sandy tweets. The disaster resources I share are a part of a very small minority.

Many of the Sandy tweets were not focused on resource sharing. Instead, many were focused on sharing a personal perspective through humor, well wishes, or updates on that user’s specific situation. These tweets may be useful in their own right as they serve to inform people who know the account owner of that person’s welfare, help people feel connected to those in disaster areas, or serve as a stress release.

Hurricane Sandy tweets included many different hashtags (#sandy, #frankenstorm, #superstorm, etc.), and others had no hashtags, making it difficult to compile the tweets and conceptualize the amount of hurricane–related tweeting that took place. The localized disasters I focus on usually have only one or two hashtags that are consistently added to most tweets (e.g. #fourmilefire). This means that the content I produce is more likely to be compiled alongside most tweets related to that disaster. However, if I am tweeting about a disaster that is being talked about nationwide or worldwide, using multiple hashtags may be a way to increase the visibility of my tweets. Other ways to increase visibility might include tweeting the same information more than once, at different times of day, or mentioning those I know in the affected area in my tweets. For example, my tweet with the #sandynyc hashtag sharing the link to a Google crisis map for the New York City area [8] may have reached a wider audience if I had included more general hashtags, such as #sandy or #hurricane, as well as the location–specific hashtag. I also could have mentioned or direct messaged specific users if the information seemed especially important to share with them (e.g., if they lived or worked in an area listed as an evacuation zone).

These findings have additional implications for my behavior as a Twitter user. I usually tweet information at the moment I locate it, which is valuable for a fire because fires can change direction, expand, or be contained in relatively brief periods of time. However, for something slow moving (like a hurricane), it may be more valuable for those using Twitter as an information resource if I stagger the tweets at different times so they appear in the stream at several intervals.

In our study, dividing up our research into four disaster management phases allowed us to see the changes in content being shared over time. This approach can be applied to tweeting to help me better tailor information to disaster survivors’ greatest needs at any given moment. It may be helpful to initially offer information on shelters for evacuees prior to a disaster hitting or while it is taking place, then offering resources on recovery funding after the danger has passed. As a content producer, it may also be helpful for me to examine what else is being shared under the hashtag(s) I plan to use to understand those information flows within a particular moment, to see how many tweets my disaster tweets will be competing with for attention. With natural disasters, this appraisal process can be ongoing, as new content becomes available and relevant (e.g., new evacuation boundaries or updated shelter listings, followed by news stories or Web sites to donate to disaster victims after the fact).

Utz, et al. (2013) found that people still trust traditional information sources, even if they use social media. This suggests to me as a Twitter user the importance of sharing a variety of sources that speak to different appraisals of value. These could include some that have value (such as Google Crisis Maps) but may not be seen as valuable to all users, to sources from news outlets that provide helpful information and may be more readily accepted by other users. In the tweets I looked at, I also found that the external information (those tweets containing a brief synopsis linking to a URL containing longer text) being reshared most frequently tended to be from news sites. As a Twitter user, this underlines the importance of sharing resources that match the needs of other users, and the importance of sharing information that will be valued highly if I am interested in having that information disseminated widely.

 

++++++++++

Thinking from both sides: How does this approach affect future research?

As a researcher, this study and the subsequent self–reflection process have a variety of applications. First, it has the potential to encourage other researchers to engage in this process of reflective research, and can inform research design. For example, one can compare content analysis of a small subset of individual tweets to the content of a much larger sample. Our study taught me the need for large samples to uncover trends, even when not conducting a quantitative study. Our preliminary analysis has only scratched the surface, and opens the door for the coding of larger amounts of data in order to draw more well–informed conclusions about the trends we see. It also reiterated the importance of qualitative coding approaches: Some tweets were ambiguous or subtle in a way that coding by keyword might not have captured. This demonstrates the complexity of the Twitter environment and the crossover between what seem to be discrete categories (e.g., humor and requests for assistance).

These interactions are found in existing research as well. Ahn (2013) argued that social media and the analysis of data inform each other, with social media serving as a knowledge source that can be understood through analysis. This played out in Ahn, et al.’s (2012) study on the Encyclopedia of Life (EOL), a crowdsourced citizen science project, where researchers found that the introduction of new features increased member participation. Additionally, they found that curators (members who were particularly active, given special privileges, and considered experts in EOL space) could encourage the participation of others by commenting on their posts and administering content. They used static visualizations to analyze engagement, but also used dynamic network visualization using a tool called TempoVis (Ahn, et al., 2011) that allowed them to track the changes in engagement between users over time, and envision how different users affected others’ involvement in that particular online space.

These findings have implications for disaster tweeting research, as well as the practical use of Twitter as a tool during disasters. First of all, studies based in similar research questions, that ask both how the tool itself impacts engagement, and how other users impact engagement, would allow researchers to track the effectiveness of Twitter as a disaster information tool from multiple angles. In this case, seeing what features of Twitter encourage engagement (retweets, favorites, etc.) may be more effective than looking at new features, as was done in the EOL study, since the features offered by Twitter are not added to regularly.

When looking at users’ effect on each other, social network analysis should center on active users’ tweets and the amount of engagement surrounding those — retweets, favorites, replies, etc. In our ongoing analysis of Hurricane Sandy tweets, we have found numerous active users who have been retweeted many times, sharing a variety of content types — humor, information to help disaster victims, well wishes for those in affected areas. Future analysis could build upon this to uncover which content types that are most likely to encourage engagement, and thus which ones are the most effective for sharing information during a disaster. It could also use a tool like TempoVis to see how those interactions change over time. This could be taken even further by asking the most popular or active content creators to self–assess their social media experiences through surveys and/or interviews. These findings can also be extrapolated to organizational Twitter use. For example, Larsson and Ågerfalk (2013) found that, while the Swedish train operator SJ used social media during a period of high travel and unexpectedly heavy snowfall, its Twitter usage still followed off–line patterns — tweets only being produced during regular workday hours, despite the trains running a 24/7 schedule. Studies that highlight the subjective experience of the user alongside evaluations of effective practice could help organizations improve social media communications, allowing them to update patrons more quickly while increasing customer satisfaction.

 

++++++++++

Conclusion

This paper has outlined how research on disaster tweeting informs the act of disaster tweeting as a user, and how those user behaviors inform my activities as a researcher. By looking at user engagement and research as informing each other, it becomes possible to create vibrant research, engage in thoughtful reflection about personal social media use, and more effectively use and research social media tools for information dissemination during a natural disaster.

Recent research has indicated that Twitter users may not be representative of the population as a whole (Mitchell and Hitlin, 2013), and that users value and trust certain types of shared content more than others (Utz, et al., 2013). This underscores the importance of tailoring content to the users who are there, rather than to one’s own perception of the environment. Being aware of research already conducted on disaster tweeting is an excellent way to do this. Additionally, it emphasizes the need for research that takes into account the user experience and the complexity of social media spaces, in order to understand and encourage engagement, create increasingly effective tools, and use existing tools most efficiently. End of article

 

About the author

Julia Skinner is a doctoral student in information studies at the Florida State University School of Library and Information Studies. Her research interests include examining how users engage with social media tools to create change. Skinner is originally from Boulder, Colo. and has her Master’s in Library & Information Science and graduate certificate from the Center for the Book in the University of Iowa.
E–mail: JuliaCSkinner [at] gmail [dot] com

 

Acknowledgements

The author thanks Jonathan Hollister, Wei Qiang, Eric Camil, and Vanessa Dennen of Florida State University for all their diligent and ongoing work.

 

Notes

1. E.g., Bunce, et al., 2011; Bruns and Liang, 2012; Freeman, 2011; Jung, 2012; and, Pew, 2012.

2. An older, yet still frequently cited, example is Morgan and Smircich (1980).

3. E.g., Akerlof, et al.’s (2013) work on perceptions of climate change.

4. In order to protect the identities of Twitter users, I have not pulled full tweets from my subjects verbatim. Instead, I have kept the text of my own tweets intact but have created representative messages in the case of others’ tweets. More information on the reasons behind this approach, and engaging in this practice rigorously and thoughtfully, can be found in Markham (2012).

5. J. Skinner, tweet on 27 June 2012, 7:05 AM.

6. J. Skinner, tweet on 27 June 2012, 7:09 AM.

7. J. Skinner, tweet on 26 June 2012, 8:14 PM.

8. J. Skinner, tweet on 28 October 2012, 2:44 PM.

 

References

J.–W. Ahn, 2013. “Data analysis and social media: How to help it and get helped,” presentation at Florida State University, School of Library & Information Studies (28 March).

J–W. Ahn, J. Hammock, C. Parr, J. Preece, B. Shneiderman, K. Schulz, D. Hansen, D. Rotman, and Y. He, 2012. “Visually exploring social participation in Encyclopedia of Life,” paper presented at the ASE International Conference on Social Informatics, at http://www.cs.umd.edu/~ben/papers/Ahn2012Visually.pdf.

J–W. Ahn, M. Taieb–Maimon, A. Sopan, C. Plaisant, and B. Shneiderman, 2011. “Temporal visualization of social network dynamics: Prototypes for nation of neighbors,” In: J. Salerno, S.J. Yang, D. Nau, and S.–K. Chai (editors). Social computing, behavioral–cultural modeling and prediction. Lecture Notes in Computer Science, volume 7227. Heidelberg: Springer, pp. 309–316.

K. Akerlof, E.W. Maibach, D. Fitzgerald, A.Y. Cedeno, and A. Neuman, 2013. “Do people ‘personally experience’ global warming, and if so how, and does it matter?” Global Environmental Change, volume 23, number 1, pp. 81–91.
http://dx.doi.org/10.1016/j.gloenvcha.2012.07.006, accessed 22 August 2013.

C.B. Brettell (editor), 1993. When they read what we write: The politics of ethnography. Westport, Conn: Bergin & Garvey.

I. Condry, 2011. “Post–3/11 Japan and the radical recontextualization of value: Music, social media, and end–around strategies for cultural action,” International Journal of Japanese Sociology, volume 20, number 1, pp. 4–17.
http://dx.doi.org/10.1111/j.1475-6781.2011.01144.x, accessed 22 August 2013.

M. Freeman, 2011. “Fire, wind and water: Social networks in natural disasters,” Journal of Cases on Information Technology, volume 13, number 2, pp. 69–79.
http://dx.doi.org/10.4018/jcit.2011040105, accessed 22 August 2013.

H. Gao, G. Barbier, and R. Goolsby, 2011. “Harnessing the crowdsourcing power of social media for disaster relief,” IEEE Intelligent Systems, volume 26, number 3, pp. 10–14.
http://dx.doi.org/10.1109/MIS.2011.52, accessed 22 August 2013.

L. Hjorth and K.–H. Kim, 2011. “Good grief: The role of social mobile media in the 3.11 earthquake disaster in Japan,” Digital Creativity, volume 22, number 3, pp. 187–199.
http://dx.doi.org/10.1080/14626268.2011.604640, accessed 22 August 2013.

L. Krefting, 1991. “Rigor in qualitative research: The assessment of trustworthiness,” American Journal of Occupational Therapy, volume 45, number 3, pp. 214–222.
http://dx.doi.org/10.5014/ajot.45.3.214, accessed 22 August 2013.

A.O. Larsson and P.J. Ågerfalk, 2013. “Snowing, freezing ... tweeting? Organizational Twitter use during crisis,” First Monday, volume 18, number 6, at http://firstmonday.org/article/view/4294/3688, accessed 7 August 2013.
http://dx.doi.org/10.5210/fm.v18i6.4294, accessed 22 August 2013.

M. Latonero and I. Shklovski, 2011. “Emergency management, Twitter, and social media evangelism,” International Journal of Information Systems for Crisis Response and Management, volume 3, number 4, pp. 67–86.
http://dx.doi.org/10.4018/jiscrm.2011100101, accessed 22 August 2013.

K.H. Leetaru, S. Wang, G. Cao, A. Padmanabhan, and E. Shook, 2013. “Mapping the global Twitter heartbeat: The geography of Twitter,” First Monday, volume 18, number 5, at http://firstmonday.org/article/view/4366/3654, accessed 7 August 2013.
http://dx.doi.org/10.5210/fm.v18i5.4366, accessed 22 August 2013.

A. Markham, 2012. “Fabrication as ethical practice: Qualitative inquiry in ambiguous Internet contexts,” Information, Communication & Society, volume 15, number 3, pp. 334–353.
http://dx.doi.org/10.1080/1369118X.2011.641993, accessed 22 August 2013.

A. Mills, R. Chen, J. Lee, and H.R. Rao, 2009. “Web 2.0 emergency applications: How useful can Twitter be for emergency response?” Journal of Information Privacy & Security, volume 5, number 3, pp. 3–26.

A. Mitchell and P. Hitlin, 2013. “Twitter reaction to events often at odds with overall public opinion,” Pew Research Center (4 March), at http://www.pewresearch.org/2013/03/04/twitter-reaction-to-events-often-at-odds-with-overall-public-opinion/, accessed 20 March 2013.

G. Morgan and L. Smircich, 1980. “The case for qualitative research,” Academy of Management Review, volume 5, number 4, pp. 491–500.

Pew Research Center’s Project for Excellence in Journalism, 2012. “Hurricane Sandy and Twitter,” at http://www.journalism.org/index_report/hurricane_sandy_and_twitter, accessed 15 December 2012.

E.M. Rogers, 1995. Diffusion of innovations. Fourth edition. New York: Free Press.

Twitter, 2011. “#numbers,&Rdquo; at http://blog.twitter.com/2011/03/numbers.html, accessed 21 February 2013.

S. Utz, F. Schultz, and S. Glocka, 2013. “Crisis communication online: How medium, crisis type and emotions affected public reactions in the Fukushima Daiichi nuclear disaster,” Public Relations Review, volume 39, number 1, pp. 40–46.
http://dx.doi.org/10.1016/j.pubrev.2012.09.010, accessed 22 August 2013.

 


Editorial history

Received 29 March 2013; revised 7 August 2013; accepted 20 August 2013.


Creative Commons License
This paper is licensed under a Creative Commons Attribution–NonCommercial 3.0 United States License.

Natural disasters and Twitter: Thinking from both sides of the tweet
by Julia Skinner.
First Monday, Volume 18, Number 9 - 2 September 2013
http://www.firstmonday.org/ojs/index.php/fm/article/view/4650/3741
doi:10.5210/fm.v18i9.4650





A Great Cities Initiative of the University of Illinois at Chicago University Library.

© First Monday, 1995-2017. ISSN 1396-0466.