By Michael Martinez and Lori Cameron
Twitter may amount to nothing more than 140 characters, but the social media channel inspires serious research that goes thousands of words deeper to address the biggest issues of our day: extremism, jihad, gangs, misinformation, and, of course, how to become the next famous influencer.
Those topics are just some of the subjects that Computer Society researchers have tackled in recent years when writing about new media and its titan called Twitter.
If you have ever wondered how social media intersects with such serious matters in our lives, here are recent peer-reviewed studies into the micro-blogging platform. They are culled from the Computer Society Digital Library (Abstracts of the research are free, but access to full-length version requires a log-in.) You may never read a tweet the same way again.
Hunt for influence maximization algorithm
Let’s go to the obvious and easy stuff, first.
Simply put, it’s what everyone wants on social media: Influence.
After all, that’s why we post.
So what’s among the next big things for Twitter?
It’s “influence maximization.”
This is where all the parties interested in making money on social media — the influencers, the marketers, the big corporations seeking to sell you a product — are trying to devise the ideal model for getting the biggest bang for their buck.
In short, Big Business is trying to identify what fixed number of influencers — let’s say 10 or fewer — can sway the biggest numbers of audiences on Twitter.
It’s good business if you can find that sweet spot, under budget, for a product promotion.
Current models for measuring social media influence totally ignore the long-term influence effects of dynamic interactions, so researchers are developing a better model that includes that interaction data for better results.
They call it a “temporal influence model,” and they contend it’s the first of its kind.
The algorithm is designed to predict future opinions of social media users, especially those users with a high diversity of opinions.
“To the best of our knowledge, our work is the first to model interpersonal influence as the continuous impact of opinion behaviors using real-world social media data,” say authors Chengyao Chen and Wenjie Li of Hong Kong Polytechnic University, Dehong Gao of Alibaba Group, and Yuexian Hou of Tianjin University.
Hunting against jihad
Researchers in Sweden, Austria, and London have devised a machine learning approach to automatically detect messages released by jihadist groups like ISIS on Twitter.
The effort is designed to help human analysts who now must plow through assorted Twitter feeds for such violence-inciting content. Once found, those Twitter accounts are suspects.
“Even though our results are preliminary and more tests needs to be carried out we believe that results indicate that an automated approach to aid analysts in their work with detecting radical content on social media is a promising way forward. It should be noted that an automatic approach to detect radical content should only be used as a support tool for human analysts in their work,” say authors Michael Ashcroft of Uppsala University in Sweden, Ali Fisher of VORTEX at the University of Vienna, Lisa Kaati of Uppsala University, Enghin Omer of Uppsala University, and Nico Prucha of King’s College in London.
All their experiments were carried out on Weka, which is a suite of machine learning software written in Java.
At an intersection with street gangs
Yes, even thugs in the ‘hood use Twitter, intimidating rivals or bystanders, representin’ with hostile imagery.
The challenge for law enforcement and algorithm engineers is to find gangs’ profiles among the hundreds of millions of users on Twitter.
Researchers at Wright State University in Dayton, Ohio, outlined a process “to curate one of the largest sets of verifiable gang member profiles that have ever been studied.”
Their search involved the study of profane language (specified within the research) and established differences in the language, YouTube links, and even emojis that gang members use compared to the rest of the Twitter population.
“Despite the challenges in developing such automated systems, mainly due to difficulties in finding online gang member profiles for developing training datasets, we proposed an approach that uses features extracted from textual descriptions, emojis, images and videos shared on Twitter (textual features extracted from images, and videos),” the authors wrote.
Illicit drug use, too
Illicit opioid and heroine abuse are among the epidemics of our day, and Twitter provides insight into the attitudes and behaviors of the sellers and buyers of illegal drugs.
“We find that prescription drugs, such as sleeping pills, Xanax, and Adderall, are the top drugs needed in Twitter,” researchers said in a 2016 study.
“Significantly more users are trying to sell illegal hard drugs, such as heroin and cocaine, rather than buy or claim need for these drugs. Finally, the Twitter chatter related to specific drugs is directly impacted by high-visibility media events involving such substances,” researchers Iris Seaman and Christophe Giraud-Carrier of Brigham Young University said.