During the week of April 15-19, 2013, two critical events turned the focus of the nation back to domestic terrorism after a nearly decade-long hiatus. On Monday, April 15, two improvised explosive devices went off near the finish line of the Boston Marathon, killing three spectators and injuring more than 200 others. This initial event resulted in a week long investigation, the identification of two suspects, the killing of a police officer at MIT, and the lockdown of an entire city. As the nation watched broadcast news, social media served as an additional channel to relay and receive time and safety critical information. This research report will provide results from analyses on the uses of Twitter by official organizations over the course of the week.
Online Response by National, State, Regional, and Local Selected Accounts
We selected 29 local, state, and federal accounts that were relevant to the Boston community and the emergency response. These focused accounts range from emergency responder and public safety organizations such as “Boston_EMS”, “Boston_Police” and “MassEMA”, to public officials at the state and local level, such as “mayortommenino” and “MassGovernor”. We also included the official alert and notification account for the City of Boston “AlertBoston,” and the Boston Police PIO, “CherylFiandaca.”
We begin our analysis by considering the rate of posting for these accounts of interest. In the top panel of Figure 1 below, we show the estimated posting rate per hour summed across the selected accounts. Prior to the marathon bombing, daily posting rates were consistent but low. Significant increases in posts are seen on the day of the marathon (April 15), the day following the marathon (April 16) and at the end of the week, when the city of Boston sheltered in place during the manhunt in Watertown (April 18-19).
Figure 1. Tweeting Network of Selected Accounts and Posting Rates for April 15-19, 2013.
We observe strong seasonality effects across the weeklong event, as seen in the significant decrease in tweet volume during the nighttime hours and the increase in the daytime hours. We note one exception to this. There is a significant spike in the number of tweets on the morning of April 19 when residents in the City of Boston received messages advising them to shelter in place. These early morning official tweets on April 19 also had some of the highest retweet rates.
The network of Twitter following relationships among our selected accounts is shown in the bottom panel of Figure 1. In this visualization each selected account is represented by a node (circle). We show the relationships between each node as grey arrows, demonstrating who is following whom. The direction of these arrows indicates the direction of attention. These following relationships could be dynamic over the period of observation, however, we find they remain stable — these organizations do not make any new connections between each other. The size of each node shows the number of followers that each account has at any given time. Nodes that grow in size experience large increases in their follower count over the week. This figure also shows activity. Each time a tweet is sent the node is shown in red; ties to that actor’s followers are shown in blue, demonstrating how information flows through this network. We also highlight specific messages; nodes are shown in green if their tweet received more than 450 retweets. Tweets with high numbers of retweets have greater exposure.
Changes in Follower Numbers
Next we consider the effect of the two events on the number of followers for these accounts. On the day of the marathon (4/15), we saw sharp increases in the numbers of followers relative to initial follower counts, with the most significant gains made by Boston Police (more than 60,000 followers added) and Boston Police PIO CherylFiandaca (adding 6,438), both representing local law enforcement. A second increase in followers is made at the end of the week (4/19) when Boston_Police adds more than 188,000 followers, MassStatePolice gains 17,918, and Boston MayorTomMenino increases his following by 13,640.
In Figure 2, we show the percent change in follower counts beginning with the week prior to the Marathon, through the end of the month. The first panel, representing local accounts, shows the greatest increases in followers (up to 45%); the second panel, representing state level accounts, shows moderate increases (up to 12.5%); and the third panel, representing federal level accounts, shows the lowest percent change. The bands are ordered alphabetically, and cumulatively the stacked bands represent the sum change across all accounts in that panel. Individual differences are seen in the width of each colored ribbon. The greatest percent change is seen at the local level, where the Boston Police and Boston Police PIO, Cheryl Fiandaca increased by more than 500% and 2000%, respectively, over the course of the week.
Figure 2. Percent Change in Follower Counts April 14-April 28.
Here are the raw numbers for the selected accounts that had the greatest increases in their follower counts over the weeklong period:
In Figure 3, we show the changes in follower counts for ten of our selected accounts that had the greatest increases (ranked by percent change). Most important here are the days on which the new followers were added in relation to the events occurring at the time. For all of these accounts, we observe very stable follower counts for the days prior to the marathon. We then see increases on two days, the day of the marathon, and then on the day of the manhunt. Following the manhunt, we see that the gains made by our selected accounts hold steady and there is little to no decline in the weeks following.
A few accounts stand out as they gained large numbers of additional followers during pivotal moments during the week. Cheryl Fiandaca’s account started with around 300 followers prior to the events that occurred on the day of the marathon. On the day of the marathon, she gains nearly 6,500 new followers and tweets consistently throughout that day. Following that initial spike, she posted no additional tweets from this account and gained very few followers throughout the remainder of the week. The Boston_Police account also experiences significant growth on the day of the marathon (more than 60,000 new followers), and there are consistent posts from this account throughout the day. This account continues to gain followers through the week at a steady pace, until the night of the shooting at MIT and the following day, when more than 80 tweets were posted over a 36 hour period and the entire city of Boston was on alert. One other account experiences significant follower gains, but has a notable lack of twitter activity. This is BostonFire, which explained its lack of presence during the emergency response to the marathon bombing by stating “we were on scene but it is our policy to not knowingly show any injured person or discuss our treatment to them.”
The increased follower counts are public demonstrations of account salience during this weeklong event. This may indicate that specific organizations are perceived as more relevant in relation to the events of interest. For this particular hazard event, local organizations appear to be most highly valued for their information, especially those that represent law enforcement and local government. Furthermore the volume of new followers added at different points along the week is an indicator of increased attention being directed to social media and the Twittersphere. Early attention to local accounts, in the immediate hours following the bombing, shows heightened awareness and information seeking in the aftermath of a local attack. During the latter part of the week, when a large portion of the Boston public was identified to be at risk and therefore placed on lockdown, growth in account followers was again witnessed at a substantially increased rate.
Importantly, we note that the pre-event posting rate and numbers of followers for this set of selected accounts was very consistent. This is also the same post event. There are no indications that followers gained will be lost in the immediate aftermath of this event, though over time it is likely that these actors will experience a decline in the follower count. This short term increase in public attention is both a risk and an opportunity. The public is still connected to these official accounts and organizations have the opportunity to develop relationships, demonstrate their continued value, and even provide educational information that may be useful for future hazard or disaster events. This also means that organizations that have increased their network size must provide information of value and to be aware that the public is watching.
This material is based on research supported by the National Science Foundation under awards CMMI-1031853 and CMMI-1031779 and by the Office of Naval Research under award N00014-08-1-1015.
The above analysis was done by HEROIC PIs Jeannette Sutton and Carter Butts and team members Emma Spiro and Britta Johnson. Please cite as follows.
Sutton, J., Spiro, E., Johnson, B., and Butts, C. (2013). “Following the Bombing.” Online Research Highlight. http://heroicproject.org