Scientists simulated a new social media platform populated entirely by artificial intelligence chatbot users and found that it led to “worsened outcomes” with amplification of polarised voices.
Social media platforms like X and Facebook have been widely criticised for promoting polarising content and for providing limited space for constructive political dialogue.
Internet safety researchers have warned that such platforms amplify sensational content and turn into echo chambers, insulating users from opposing perspectives.
A growing body of studies has examined links between algorithms driving these social media platforms and the polarisation of content posted on them.
“Platform algorithms – optimised to maximise user engagement – often have the unintended effect of amplifying outrage, conflict, and sensationalism,” researchers write in a yet-to-be peer-reviewed study, posted in arXiv.
In the study, researchers tested a social media platform with AI users powered by ChatGPT-4o to see if they could stop it from turning into an echo chamber for its artificial userbase.
They tested several strategies, including switching the content feed to a chronological order, intentionally boosting diverse viewpoints, removing account bios and even hiding user stats like follower counts as means to stop the platform from turning into an echo chamber.
“Our goal is to simulate a stylised social media environment to assess whether it reproduces key dysfunctions identified in the literature – such as polarisation, attention inequality, and engagement-driven distortion,” scientists wrote.
They then tested whether their interventions mitigated these dysfunctions.
However, researchers found that these strategies only had “modest improvements – and in some cases, worsened outcomes”.
“While several showed moderate positive effects, none fully addressed the core pathologies, and improvements in one dimension often came at the cost of worsening another,” they noted.
“The results are sobering: improvements are modest, no intervention fully disrupts the mechanisms driving these outcomes, and some changes worsen the problems they aim to solve,” the scientists write.
The findings suggest social media platforms could be doomed to become highly polarised even in the absence of recommendation algorithms or engagement optimisation.
Researchers suspect any meaningful reform of social media platforms would require a “fundamental redesign”.
“This study has demonstrated that key dysfunctions of social media can arise even in a minimal simulated environment that includes only posting, reposting, and following, in the absence of recommendation algorithms or engagement optimisation,” researchers wrote.
“The emergence of these properties from a minimal platform suggests that these problems may be rooted not in the details of platform implementation or algorithms, but in deeper structural mechanisms.”
Google Translate adds AI with speed and accuracy modes, Duolingo-style practice
AI chatbot can now end conversations if it is distressed
Humanoid robots are on sale for £4,400. Here’s what it can do
Groundbreaking brain chip decodes people’s inner monologues in real time
US state claims TikTok uses addictive algorithms to target children
UK ‘agreed to drop’ Apple data demand in privacy row, US chief says
