Research in the International Journal of Computational Science and Engineering has developed a new approach to addressing ideological polarisation on social media. The problem of users generally encountering only like-minded perspectives and so reinforcing their own beliefs even in the face of conflicting evidence is highly divisive.
The phenomenon, known as the “echo chamber” effect or referred to as “filter bubbles”, arises in part because the algorithms driving the position of content in one’s social media apps. This, in turn, is driven largely by the need to keep users active and engaged on a particular platform. Too many contrary updates might drive users away, and that will ultimately reflect negatively on the advertising and other revenue streams for the companies that operate the platforms. By contrast, an echo chamber effect that reinforces their viewpoints will, for many people, be more attractive than one that doesn’t.
Zaka Ul Mustafa and Muhammad Amir of the International Islamic University Islamabad, Manal Mustafa of Zaman Technologies Pvt Limited, Pakistan, and Muhammad Adnan Anwar of Ulisboa, Portugal, suggest that the social media platforms could benefit from the use of genetic algorithms (GAs). Such computational techniques inspired by the principles of evolutionary natural selection could reduce polarisation and the echo chamber effect but still respect the organic nature of online interactions, and so keep users engaged without being so divisive.
The team explains that current strategies to counter polarisation often involve connecting disparate groups (edge addition) or altering expressed views (opinion flipping). These methods are not only static, but also raise ethical concerns about the platforms interfering with user autonomy. A GA-based approach instead identifies influential nodes in the online social network and only subtly adjusts their highlighted connections to reduce polarisation. The critical contribution of the work lies in identifying network elements that disproportionately contribute to ideological divides, and then encouraging more diversity of interaction with minimal disruption to the organic nature of social media.
The team has tested their approach on real-world datasets that focus on polarised US political discourse. The datasets have communities clustered around distinct ideological groups, and so can provide a useful test for how well the method precludes polarisation and division. The results showed that the GA approach could foster connections between disparate groups, and this led to a measurable decrease in polarisation without fundamentally altering the network’s overall structure.
Ul Mustafa, Z., Amir, M., Mustafa, M. and Anwar, M.A. (2025) ‘Harmony amidst division: leveraging genetic algorithms to counteract polarisation in online platforms’, Int. J. Computational Science and Engineering, Vol. 28, No. 7, pp.1–17.