News of the Westminster attack on March 22 spread the way many terror attacks do: an eyewitness tweet, BBC live coverage and lots of aerial video from helicopters while news outlets scrambled to put the pieces together. London watched in shock as Big Ben became the center of a high-alert police cordon. But before they could mourn, many simply wanted to know what happened.
As is the case with many breaking news events, Twitter became one of the quickest ways to see up-to-the-second developments in the story. To better understand how people use the social network in times of chaos, I sifted through a sample of 50,000 tweets scraped over a 30 minute time period. These tweets were sent just one hour after the attacker began his deadly drive across Westminster Bridge.
What I found was a vast mixture of sentiments: everything from solidarity and crowdsourcing to the danger of Islam and immigrants. After running some statistical analysis, the following network graph shows that eight potential communities can be found by mapping the retweet pattern of all tweets with #Westminster from this time period.
Each color on the graphic represents a different community of tweets relating to the attack. A node can connect to two different colors, but is colored based on its original source tweet. The lines represents retweets of original tweets sent by one account. Though many accounts retweeted many times over the course of the day, here are a few takeaways I found when analysing the data.
Most people wanted to help or show solidarity
Of the communities found in the dataset, at least three of them were related to crowdsourcing or showing solidarity. @MetpoliceUK was one of the most widely retweeted accounts and represents the bulk of all pink and green colored nodes. The most widely retweeted tweet from this time period was from the Met, urging people to show caution and respect for the victims.
The Met also asked people to send them any information about the event. This tweet was widely circulated, regardless of the community in the graph.
The gray community represents non-news accounts and generally contains tweets showing concern or solidarity with the victims of the attack. However this group of sentiments is likely much larger than this one color depicted in the graph, since all tweets in French and Spanish were automatically given a separate category due to difference in language.
Large news networks are still the most shared source in chaos
Though Twitter allows for anyone to spread news of an event such as this, the large news networks such as @BBCBreaking and @AP are still the go-to sources for news coverage of unfolding events. So much so that these tweets were categorized in a distinct community.
Other news networks quickly covered the story as well (represented in orange). Accounts like @DailyMail, @Skynews, @iTVNews, @DailyMirror, and @Telegraph are spread through the center of the network, acting as the base on which other communities grow.
Far-right, nationalist accounts jump on the opportunity to spread fear
Where there is fear, there will always be people willing to stoke it. The community represented in blue at the bottom of the graph can generally be described as expressing far-right or nationalistic sentiments. These tweets range in severity and ideology. Accounts with the widest reach in retweets include @DefendEvropa, @V_of_Europe and @Foxnews.
Some tweets were simply spreading the news of the attack, or urging caution. But some of the largest nodes in the set (and therefore at the time, the most retweeted) were tweets with anti-Islamic sentiment.
Whether overtly political or just a gut reaction to the event, the tweets found in the blue community seem to express a strong nationalistic ideology. These tweets remain fairly isolated in their community but in some instances, spread to other parts of the graph.
When accounting for the fact that two communities are dominated by @MetpoliceUK tweets, the far-right community ranks as the second highest in the amount of related retweets.
Though these eight communities emerged from the dataset, more analysis is needed to see how messages are spread over time. However, as a sample of tweets taken just one hour after the attack when many questions were still left unanswered, the data still reveals some insights into how messages are spread on Twitter in times of chaos and what types of accounts are sharing them.
Tools used for analysis:
- Python: for scraping Twitter
- Excel: for cleaning data
- Gephi: for creating network graph and running statistics
- Adobe Illustrator: for creating graphics