Jeremy Blackburn is a computer scientist who is part of a Binghamton University research team that has developed machine learning algorithms that can successfully identify bullies and aggressors on Twitter with 90% accuracy. While effective tools for detecting harmful actions on social media are rare, this team of researchers is creating one that’s correct almost every time! The research has been published on Researchgate.
The reason such tools are scarce is that bullying behavior is frequently ambiguous. Furthermore, it is often exhibited through seemingly superficial criticisms and comments. To bridge this gap, the research team analyzed the behavior patterns displayed by abusive Twitter users and compared them with other Twitter users to find their differences.
Blackburn said:
We built crawlers — programs that collect data from Twitter via a variety of mechanisms. We gathered tweets of Twitter users, their profiles, as well as (social) network-related things, like who they follow and who follows them.
After that, the team carried out natural language processing, as well as sentiment analysis on the tweets themselves. They also conducted several social network analyses on the connections between users.
From all the data collected, the team developed algorithms that could automatically classify two particular types of offensive online behavior: cyberaggression and cyberbullying.
They put their algorithms to the test and found they were able to classify abusive users on Twitter with an accuracy of 90%! By abusive users, they mean those engaging in harassing behavior, such as making racist remarks or sending death threats.
Blackburn said:
In a nutshell, the algorithms ‘learn’ how to tell the difference between bullies and typical users by weighing certain features as they are shown more examples.
As lovely as this development is, it’s still just the beginning of what they hope to accomplish. The research has only helped one aspect of the problem, which is mitigating cyberbullying. There’s always the problem of how the bullying affects the user it was targeting.
Blackburn said:
One of the biggest issues with cyber safety problems is the damage being done is to humans and is very difficult to ‘undo’. For example, our research indicates that machine learning can be used to automatically detect users that are cyberbullies, and thus could help Twitter and other social media platforms remove problematic users. However, such a system is ultimately reactive: it does not inherently prevent bullying actions, it just identifies them taking place at scale. And the unfortunate truth is that even if bullying accounts are deleted, even if all their previous attacks are deleted, the victims still saw and were potentially affected by them.
The researchers are now investigating pro-active mitigation techniques to manage harassment campaigns.
