A new study has revealed that any effort to eliminate users’ behavior through the algorithm will be unsuccessful. It seems that the toxicity of social networks can no longer be corrected.
O Polarizing impact of social networks It is not just the result of algorithms that we consider bad. It is now inevitable thanks to the central components of how platforms work – and will continue to work.
The conclusion is from a new new study, last week in the arXivthat Ilibates the algorithm for toxicity in the pitty networks.
500 AI chatbots were created to imitate a variety of political beliefs in the US, based on American National Election Studies Survey. These bots, fueled by the GPT-4O Mini Language Language model, were instructed to interact with each other on a simple social network that researchers had conceived without ads or algorithms.
For five phases of the experience, each involving 10,000 actions, the AI agents always tended to follow people with whom they shared political affiliationswhile those with more party opinions gained more followers and republications. This reflected the general attention to these users, which gravitated towards more party authors.
“Poisoned tree fruit”
“We were waiting for polarization to be driven by algorithms. We thought the platforms would be designed for this-to produce these results-because they are designed to maximize involvement and irritate us, and so on,” he told, to,, Petter Törnbergfrom the University of Amsterdam, which led the investigation.
After all, it seems not. The study suggests that It is not the algorithms that are causing the problemwhich can make any attempt to eliminate the antagonistic behavior of users very difficult through design.
“Most activities on social networks are always Poisoned tree fruit – The initial problems of social networks always reside in their fundamental design and, as such, can Encourage the worst of human behavior”, Disse, in New Scientist, Jess Maddoxfrom the University of Georgia, which was not part of the investigation.
