Twitter for crisis communication: lessons learned from Japan's tsunami disaster

2 weeks after the Great Tohoku earthquake followed by the devastating tsunami, we have sent open-ended questionnaires to a randomly selected sample of twitter users and also analyzed the tweets sent from the disaster-hit areas. We found that people in directly affected areas tend to tweet about their unsafe and uncertain situation while people in remote areas post messages to let their followers know that they are safe. Our analysis of the open-ended answers has revealed that unreliable retweets (RTs) on twitter was the biggest problem the users have faced during the disaster. Some of the solutions offered by the respondents included introducing official hashtags, limiting the number of RTs for each hashtag and adding features that allow users to trace information by maintaining anonymity. Introduction Recently there has been much interest in the use of social media in disaster communication. A wide range of studies suggest that information sharing networks, such as twitter, can be very useful in times of crisis by quickly and effectively disseminating relevant news (Vieweg et al., 2010). Most of the past research about this issue had utilized quantitative data collection methods and focused on strategic keyword detection (Vieveg et al., 2010), perception of false RTs (Mendoza et al., 2010), and categorization of the tweets sent during disasters (Sinnappan, 2010). However , there was still confusion among the public about the reliability of the tweets after the great Tohoku earthquake. This study aims to address the problems the users experienced and close the literature gap regarding effective use of twitter by qualitatively analyzing the nature of tweets and assessing the expectations of the users. Particularly, the paper seeks to address the problems experienced by the users and potential solutions that will provide actionable insights for future use of social networking tools during disasters. Background The 9.00 scale earthquake that hit Japan on March 11 at 2:45 PM (5:45 AM GMT) was not only the strongest in Japan’s history but also the 5th biggest ever recorded (Cheng, 2011). Tsunamis caused by the tremors already devastated most of the coastline in 3 prefectures and wiped out two towns off the map, claiming the lives of thousands and forcing more than 500,000 people to live in shelters. The Japanese prime minister Naoto Kan declared the incident as the biggest disaster after WWII. A month after the quake, there are 27,000 people dead or missing; not to mention 4 explosions in an earthquake hit nuclear facility and extremely high radiation levels in the surrounding areas (Bloomberg, 2011). Media reported that twitter was the only communication tool functioned properly immediately after the earthquake (Hosaka, 2011) What is Twitter? On its website, where over 95 million messages are shared every day (twitter, 2010), Twitter introduces itself as “a real-time information network that connects you (users) to the latest information about what you (they) find interesting (www.twitter.com).” Twitter has many features similar to other online social network sites such as status updates and directed social connections between users (Huberman, 2008); however, unlike other OSNs, it does not require mutual acquaintance between members for any information to be shared. Once a user posts a message on his/her twitter timeline, that message (tweet) becomes public and can be viewed by anyone. Because of this anonymity, some authors referred to twitter as an information sharing tool rather than a social network platform (Cataldi et al., 2010; Lee et al., 2010; Osch and Avital, 2010). Whether it is a social network or just a stream of short blog posts, Twitter has received a lot of attention from both industry practitioners and academicians. After its launch in 2006, twitter has grown exponentially, reaching 200 million users and recording over 30 billion messages in only 5 years (Petit, 2011). Currently, a keyword search on Google Scholar, allintittle:twitter, generates more than 3000 hits which indicates that there are more than 3000 academic papers dedicated to twitter. Studies point to the important role of twitter in terms of quickly disseminating information among publics during critical times such as Obama’s presidential election, Iran street protests (Rossman, 2010), great Chile earthquake (Mendoza et al., 2010) and Egyptian revolution in 2011 (Fahim, 2011). Some other uses of twitter can be listed as real-time earthquake detection (Sakaki et al., 2010), prediction of election results (Tumasjan et al., 2010), enterprise microblogging (Riemer, 2010) word of mouth marketing (Jansen, 2009) foreign language learning (Borau et al., 2010) and box-office performance forecasting (Asur and Huberman, 2010). Why do people Tweet? When it comes to individual twitter use, motives vary greatly with one important underlying goal: sharing information. Java et al., (2007) found that people tweeted mostly to tell others what they are doing, engage in conversations with other users and report news. After conducting a focus group with the early adopters, Mischaud (2007) came to notice that people used twitter because it allowed them to update everyone about anything, start dialogues with anyone and write diaries effortlessly. Ehrlich and Shami (2010) interviewed with IBM employees who use twitter internally and listed major motivations as a) reaching information quickly, b) gaining higher visibility, c) feeling connected and d) being aware of what is happening around. Naaman et al. (2010) also emphasized the information sharing aspect of twitter noting that the majority of twitter users tend to tweet about themselves (“meformers” versus “informers”). Twitter Usage during Disasters Recently, there has been an increasing amount of literature on how Twitter, a microblogging site, was used in emergency situations, such as the shooting of police officers(Heverin & Zach, 2010), river flooding(Vieweg et al., 2010), and grassfires(Corvey et al., 2010). Heverin and Zach(2010) analyzed the tweets posted after the shooting of four police officers in Seattle, and found that people mainly useTwitter to share information, rather than mentioning their opinions or personal emotions. Vieweg and Starbird(2010) conducted a similar study during the Red River Flooding and identified three tweet categories; namely, repeated information, linked to information on external websites, and retweets. In another study, Corvey et al.(2010)examined the effectiveness of Natural Language Processing method in identifying important keywords on twitter during the 2009 Oklahoma grassfires. The study drew attention to the problems faced by twitter users who struggled to identify important information during the disaster; however, it was concluded that because of the varying length of entities, annotators cannot be assigned simple rules. Several studies investigating the use of twitter during earthquakes have recently been carried out particularly in Chile and China. Qu et al.(2011)analyzed the use of a Chinese microblogging site, SinaWeibo, in response to the 2010 Yushu earthquake. The messages posted on Sina-Weibo were mainly categorized into 5 groups: Opinion-related (33%), situation update (25%), general earthquake related (18%), emotion-related (16%), and action-related (4%). In another study, Mendoza et al.(2010) examined the responses to true and false information that emerged on twitter after the 2010 Chile earthquake. The findings revealed that the propagation of tweets that correspond to rumors differs from tweets that spread factual news because rumors tend to be questioned and denied more than the news by the Twitter community. Mendoza and his co-authors concluded that there are measurable differences in the way messages propagate, thus it is possible to detect the credibility of tweets, with 70% to 80% accuracy.