Using tweets from 2009 to 2021, researchers created a predictive model that may identify extremist individuals and content associated with the militant organization known as the “Islamic State” (ISIS). They think their work could help social media firms find these kinds of identities faster and eventually ban them, lessening their influence on online groups.
US researchers at Pennsylvania State University recognized possible propaganda themes and their traits, and they created an image classifier to determine which image categories were most frequently associated to tweets concerning ISIS.
Younes Karimi, a doctorate candidate in informatics at the university and the lead author of a paper published in the journal Social Media Analysis and Mining, stated, “The Islamic State group and its affiliates, sympathisers, and followers continue to manipulate online communities to spread extremist propaganda.”
In order to look into the current actions of ISIS, the researchers gathered a dataset of tweets from possible ISIS sympathizers in addition to the ISIS-linked tweets they used for analysis.
Karimi claims that despite efforts by websites like as X (previously Twitter) to limit its online activity, the Islamic State organization is increasingly depending on social media to propagate propaganda, discredit its opponents, and find supporters.
In order to distinguish between people who shared content related to ISIS, the researchers employed artificial intelligence (AI) techniques, including machine learning and natural language processing. Natural language processing requires working with textual material, whereas machine learning generates predictions based on historical data.
The labeled data for the study’s ISIS users were ISIS accounts that were detected before to 2015; in order to identify possible ISIS sympathizers, the researchers used the old dataset to create a user classifier.
According to Karimi, the users in the dataset included people who retweeted, referenced, or discussed ISIS in addition to known members of the Islamic State group.
It is our belief that individuals who retweet or quote content from the Islamic State organization are more likely to be sympathizers or affiliates than those who only mention the topic. Mentioners can be seen as our non-ISIS users and non-trivial counterparts to ISIS users because their tweets are likely still connected to the group and cover themes that are comparable to those of the group, according to Karimi.
After that, the researchers examined the tweets to find what they called “candidate propaganda.” They compared the subjects covered by well-known Islamic State group accounts before 2015 in the old dataset to the posts made by possible sympathizers and affiliates after 2015 in the new dataset.
Initially, the team discovered that the content they had detected was “pervasive and continuous in the way it’s shared.” Second, the group looked at terms and visuals that promote certain ideologies, noting that these are “often designed to elicit an emotional response and influence a large audience.” Thirdly, Karimi noted that after examining hashtag-related content, “ISIS supporters and affiliates enlisted individuals to retweet hashtags in order to generate popular concepts, like heavy religious allusions, and curate group messaging to enhance the group’s branding and guarantee message longevity.” The group discovered that the most popular ISIS hashtags were
“Caliphate News,” “The Islamic State,” “Urgent,” “The State of the Caliphate,” and “ISIS.”
The dataset’s longitudinal perspective, according to Karimi, was significant since it contained information from both before and after 2015, when Twitter conducted a significant purge that resulted in the removal of user accounts and content associated with the Islamic State group.
The extremists were forced to shift their online tactics in reaction and relocate to new platforms, and since that crackdown, not much is known about their online whereabouts, according to Karimi.
