It’s a knee-jerk reaction. When a friend, a colleague, or a family member becomes radicalized into dangerous conspiratorial thinking, it’s easy to blame the dreaded algorithm of their social media outlet of choice. It started feeding them heaping spoonfuls of unhinged nonsense that’s disconnected from reality.
Brought to our attention by Gizmodo, A new study from the University of Amsterdam suggests we’ve been giving the algorithm a little too much credit and/or blame, depending on how you look at it.
It turns out we don’t need machine learning to splinter into hostile little tribes. We can do that just fine all on our own, thank you very much.
Even Without Algorithms, Social Media Still Turns Us Into Partisan Monsters, Study Finds
For the study, the researchers created 500 AI-powered chatbots, each imbued with a persona and political leaning. They were dropped into a stripped-down social media simulation.
No ads. No content recommendations. No algorithm wrapping the feed. Just bots communicating with one another via text—a classic social media clean slate.
And yet, the bots still sorted themselves into political camps to create their echo chambers. Over five experiments totaling 10,000 actions, the chatbots consistently followed users with similar views and reposted the loudest, most partisan takes.
The hottest takes drew the most attention. In our world, we tend to think it’s because the platform is manipulating engagement numbers, which does happen. However, in this case, without the invisible hand of a social media company’s algorithm silently dictating the terms of conversations, the bots in the experiment were doing it all on their own.
They were doing it because they wanted it, for as much as a non-sentient AI chatbot could “want” something. These bots were trained on data produced by actual, flesh-and-blood human social media users.
The researchers suggest that their findings are not an indictment of AI; instead, they reflect who we are. The chatbots are just playing out the scripts we’ve written with every angry tweet and tribal Reddit thread over the past two decades.
The researchers attempted to address the issue by incorporating all the usual solutions people propose when devising ways to mitigate social media’s polarization problems. They tested solutions like chronological feeds, hiding popularity metrics, and boosting opposing views.
At best, they saw a six percent shift in polarization. In some cases, it got worse when identity signifiers like bios were removed from profiles. Is it simply an issue of anonymity? Who knows. But one thing the study’s results showed for sure: it’s not just the algorithm’s fault.
Social media is just a digital ThunderDome, adding extra drama to an already dramatic debate environment. The platforms are amplifying the noise for sure, but they are just the chambers that echo our deafening and incredibly angry voices.
They wouldn’t be so bad if we weren’t so bad.
The post Even Without Bots and Algorithms, Social Media Turns Us Into Monsters appeared first on VICE.
