Going viral? Study tests the probability of social media posts achieving great success

New research from the Florida Institute of Technology can determine if a photo or video goes viral on social media.

Social media outlets like Facebook, Twitter, Instagram, and TikTok have quickly become an important and integral part — and sometimes troubling — of modern society. They are where we share our creations with the world, express our views and feelings, and stay connected with the people who matter to us.

They have also become an effective communication platform for celebrities, businesses, organizations, even governments. It is especially important for these parties to create widespread awareness among online users. So predicting whether a particular post or tweet will go viral—that is, shared among a significant number of users—is key to both smart marketing decisions and mitigation. effective misinformation/misinformation.

This summer, Xi Zhang and Akshay Aravamudan, PhD students in computer science and engineering, together with computer engineering and associate professor of science Georgios Anagnostopoulos, presented their work, “Predicting any time information cascade through self-stimulation processes,” on original online content. popular predictions at the 39th International Conference on Machine Learning (ICML), a leading conference machine learning research meeting held this year in Baltimore.

Florida Tech’s new popularity prediction scheme is based on Hawkes score processes, mathematical principles that model the timing of content sharing, such as reposts and reposts, as random times. course. The program’s processes are capable of capturing the self-stimulating nature of viral content, meaning that the program can model the “rich getting richer” phenomenon of viral content ( e.g. memes, etc.), where popular content becomes even more popular, at least for a while.

This is because, since users share it a lot online, many other users know about it and reshare it themselves. This explains the often observed Matthew effect of cumulative advantage—often referred to as the “richer get richer” effect—in social media: over a period of time a popular post became even more popular.

Zhang’s work provides a simple way to calculate the average future retweets based on the content’s popularity to date. More clearly, Zhang’s work allows for predicting how online content re-sharing will evolve over time. “Predicting the popularity of content is a challenging task,” says Zhang, “especially if done early, when the content is only recently posted and there isn’t enough initial awareness yet.”

The team built a prediction scheme that is more accurate and less computationally intensive than today’s most advanced methods.

“It’s important to be able to quickly gauge the popularity potential of a tweet, as some tweets can go viral within two or three hours,” added Zhang.

Another advantage of the new method is its interpretability. “We can make useful predictions, but we can also explain exactly why our model is predicting it,” says Aravamudan. “Being able to do so sheds new light on the mechanisms underlying the spread of popular content online.”