An experiment with agents created with Artificial Intelligence shows how a propaganda campaign can be generated to influence public opinion and the outcome of elections.
Although what appears on screen to be a real online debate during the election, the debate could be the product of Artificial Intelligence. Researchers warn that networks of AI agents could work together to coordinate disinformation campaigns during elections, flooding social media with propaganda, independently and on a large scale, and influencing public opinion.
In a recent study, scientists tested this phenomenon on a simulated social platform, similar to the X platform. They created 50 AI agents, 40 of whom acted as regular users and the other 10 as operators whose goal was to promote a fictitious candidate. They tested three scenarios: agents with a single goal; agents who knew their teammates; and agents who could plan strategies together. The main result of the experiment was that simply knowing who was on the same team was enough to create coordinated behavior, almost as fruitful as planning together.
The product was almost like real online conversations: diverse opinions, responses, and growing support around a single message. This innovation is very different from traditional bot-assisted campaigns. Older bot methods follow simple scripts: “post this, recycle that,” making them easier to detect.
AI agent systems, on the other hand, are more like real users. Although this was just a simulation, the consequences are real. Such systems could shape public opinion and increase divisiveness in upcoming elections. And they could be much harder to detect, because it’s not just about what individual accounts post, but also how networks of accounts act together. And it’s still unclear whether platforms can handle this.(DW)