Celebrating science on Twitter’s 10th birthday

Scientists have used Twitter to analyze America’s surging opioid epidemic, prevent retaliation of police brutality, and pull up politicians on their use of tweets.

To celebrate Twitter's 10-year anniversary (March 21, 2016), clement Lwe speak with social scientist Clement Levallois. The assistant professor at EMLYON Business School created a Twitter for Research handbook, borne from an international conference of researchers using the site for their science.  We ask him about some of the more interesting papers he’s come across using the social media site’s data.

ResearchGate: Why has Twitter received so much attention in the academic world?

Clement Levallois: Twitter is the only source of large scale, real time, publicly available conversations on a wide range of topics today. As such, this provides an invaluable resource to listen to the "pulse" of societies at the micro or macro level. The Handbook is a good illustration of that, with studies on cities to fandoms, on social movements and corporations. Twitter is also a social network (people follow and are followed by other users), and this is another extremely interesting layer of information. Finally, the mass of information provided by Twitter makes it an excellent testing ground for all sorts of techniques in information retrieval and predictive analytics, which are some of the main pillars of the knowledge economy.

RG: Twitter is celebrating its 10th anniversary this year – how has research on the site (or using its data) developed over this time?

Levallois: The study by Fausto and Aventurier in the first chapter of the Handbook shows that one paper on Twitter was published in 2006, and 720 papers in 2014. 2011 is the year when the count of publications really started rocketing. We can see two stages: until 2011, the few studies on Twitter were focusing on its specificity as a social media, or social network. Then starting in 2011, Twitter was studied less for itself, and more as a convenient data source to study something else - from sentiment analysis to political opinions.

RG: Where do you see Twitter research heading in the next ten years?

Levallois: It is a difficult question, and I don't pretend to have a firm answer. My guess is that Twitter has built a basis of loyal users and this core will now steadily expand. 10 years would be a reasonable time span to expect Twitter to finally become available in China, as it would increase the representativeness of the population of Twitter users. Research wise, I hope to see Twitter (and other social media) offer scientists better access to their data, especially the data on "who follows whom" which is currently very slow to harvest. This would increase the pace of fundamental research and also result in practical innovations currently held in check.

RG: Can you share the most innovative research paper you’ve seen on this topic?

Levallois: I am truly impressed by the work of the data scientist Bruno Goncalves. In recent years, Goncalves has explored the potential of Twitter to better comprehend how languages are used - at the country level and at the global level. To do that, he collects humongous collections of tweets and analyzes their lexical content in conjunction with the place and the time they were tweeted. His study identifies two well defined macroregions sharing common lexical properties for Spanish. It exemplifies how classic questions in geography and linguistics can find new answers when social media data is harnessed and used with talent.

RG: What about the most interesting or unique paper?

Levallois: It is a paper that was never published... in 2012, when the rage was all about predictive analytics based on Twitter data, I gave my students an assignment to try and predict a phenomenon of their choice. A group of students came with the result that they could beat the odds of the bookmakers on football games from the English Premier League, just by analyzing in real time the tweets posted during the games. I would be curious to see that replicated!

RG: What sets Twitter apart from other social media sites?

Levallois: Twitter is used to broadcast messages publicly, while Facebook is predominantly used to entertain conversations in the private circles of friends and family. This has very practical consequences: Twitter data can be used for research (under some precise conditions), whereas Facebook data can be used only by researchers employed by Facebook (and they do perform fascinating studies with it). Instagram has similarities with Twitter, since it is made of public feeds and enjoys a wide popularity. Today it remains easier to analyze text than images, which means that Instagram is comparatively less used than Twitter by researchers. But with great progresses currently underway in image recognition, I expect Instagram (and later, video social networks) to become very popular with researchers as well. Other social media like LinkedIn would be fascinating as well, but offer restricted access to their data at the moment.

RG: Why did you create this handbook?

Levallois: I am often surprised by the lack of communication between scientific fields, especially when they are "far apart" like the humanities and neuroscience, for example. Yet, these fields may have overlapping domains of interest and they often share methodologies or datasets, so that gains should be expected if researchers across fields would share their expertise. Twitter is a case in point. It is used as a source of data in a variety of domains but there is no venue or handbook to foster cross disciplinary communication about it. So that is how the idea came of an interdisciplinary conference on "Twitter for Research", and of a handbook to follow.


Read more Twitter-based papers...

America’s opioid epidemic prevalent on Twitter

Researchers from the University of California use Twitter to establish a link between America’s opioid epidemic and the promotion of prescription medication online. These researchers take a similar approach, analyzing the prevalence and pattern of prescription medication abuse on Twitter.

Retaliation of #CPDK (Chicago Police Department Killer)

Young black men were nine times more likely than any other American to be fatally shot by police in 2015. This researcher from Columbia University follows the reaction on Twitter after one of these incidents for early intervention and prevention strategies of retaliation.

Emotional people, emotional tweets

There’s nothing wrong with expressing emotions and tweets are no exception, but you’re wrong if you think it goes unnoticed. Researchers have dug through tweets to find signs of sentiment, and others have explored how often people are deceitful in their tweets and social media profiles.

Trump’s use Twitter a cause for concern?

Most politicians have realized Twitter is a great way to reach out to voters, but many of them need to learn the rules before they play the game. These researchers look at how (not) to talk on Twitter, and the ramifications of politicians’ ill-composed tweets.

Feature image courtesy of Nicolas Mirguet.