Scientists simulated a social media platform with only AI chatbot users and found some concerning results.
The study revealed that the platform experienced “worsened outcomes,’’ particularly with the amplification of polarized voices.
This suggests that AI chatbots interacting solely with each other can intensify extreme views and potentially create echo chambers.
Social media apps like X and Facebook have been widely criticized for promoting polarizing content and for providing limited space for constructive political dialogue.
The study has examined links between algorithms driving these social media platforms and polarization of content posted on them. However, the study is yet-to-be peer-reviewed, that has been posted in arXiv.
“Platform algorithms – optimized to maximize user engagement – often have the unintended effect of amplifying outrage, conflict, and sensationalism,” researchers write in a yet-to-be peer-reviewed study.
[2305.16941] Engagement, User Satisfaction, and the Amplification of Divisive Content on social media
Researchers have tested a social media platform with AI users powered by ChatGPT-40 to see if they could stop it from turning into an echo chamber for its artificial user-base.
The scientists have tested multiple strategies, including switching 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.
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.
The study findings suggest social media platforms could be doomed to become highly polarized even in the absence of recommendation algorithms or engagement optimization.
Despite this, researchers are skeptical that any meaningful reform of social media platforms would require a “fundamental redesign”.
This study has also showcased that key dysfunctions of social media can arise even in a minimal simulated environment that "combines only posting, reposting, and following, in the absence of recommendation algorithms or engagement optimization,” the study authors noted.