Trapped in the filter bubble: After 60 minutes, it is too late

Social media has become a central arena for public discourse and the dissemination of information. However, the personalized algorithms that curate content on platforms such as Instagram and TikTok carry the risk of filter bubble formation. In such a filter bubble, users views are reinforced in their existing opinions and worldviews because they are primarily shown content that matches their previous behaviour. This in turn poses a threat to social cohesion if individuals perceive facts primarily from individualized, possibly extreme, points of view.

SOAP produces and analyzes filter bubbles

In order to examine the emergence and speed of this phenomenon, Luka Bekavac developed the tool “SOAP” (System for Observing and Analyzing Posts) as part of his master's thesis. “Classic soap makes bubbles, SOAP makes filter bubbles,” says Bekavac about the naming. But SOAP does more than that: with the help of a multimodal language model, SOAP can analyze the content of posts in social media and thus capture the variety of topics and tonality in user feeds over time - in other words, it makes it possible to detect and measure filter bubbles.

Caught in the Trump camp within 60 minutes

The tool's initial test results are worrying: SOAP was able to create filter bubbles around the Palestine-Israel conflict in no time at all. “There was also a lot of misinformation and violent content in there, which actually violates Instagram's content guidelines,” says Bekavac, who is writing his master's thesis at the School of Computer Science at the University of St.Gallen (SCS-HSG) under Professor Simon Mayer. Bekavac also conducted some tests on the US elections. “After just 60 minutes of scrolling and liking, a user feed can be displayed in which 85 percent of the content is ‘pro-Trump’, depending on the interaction,” explains Bekavac. The research team, which, in addition to Bekavac, consists of Dr Kimberly Garcia, Jannis Strecker, Prof. Dr Simon Mayer from SCS-HSG and Prof. Dr Aurelia Tamò-Larrieux from the Faculty of Law at the University of Lausanne, has since published these and other findings in a scientific paper.

SOAP can also be used to check social media for compliance with the law.

The tool, which for the time being can only analyze Instagram content, should also be applicable to other online platforms in the future. Bekavac plans to conduct detailed evaluations of the dynamics on different social networks and on different topics as part of his doctoral thesis. “The tool is open source and should also be able to be used by other scientists to study social networks,” emphasizes Bekavac. This is because, in addition to examining filter bubbles, it is also suitable for analyzing other systemic risks of online platforms. For example, it could be used to investigate whether social media display personalized advertising to minors, which is currently prohibited. “This tool offers valuable insight into the risks of social media for accessing information and forming opinions,” says Prof. Dr Miriam Buiten, HSG expert in technology law, who co-supervised the work. ‘The EU Commission is currently developing guidelines for algorithmic recommendation systems as part of the Digital Services Act. SOAP can help to measure the impact of the changes after they are implemented.’

Self-protection against filter bubbles

It's easy to get caught in a filter bubble – and not just online, by the way. That's why it's important to be aware of it and to critically question information, especially on social media. To avoid getting stuck in your bubble, Bekavac advises using the non-personalized feed, which is possible on most platforms. On the other hand, it naturally helps if you consciously don't just follow people and like content that represents your own opinion.