arstechnica.com
A large study involving 80,000 UK participants tested the ability of 19 AI chatbots to influence political views. Researchers from institutions like the UK AI Security Institute and MIT had AI advocate for or against stances on 707 political issues, engaging participants in short conversations. Participants rated their agreement with each stance before and after interacting with the AI. While the AI chatbots didn’t demonstrate “superhuman persuasiveness,” the study explored concerns about AI’s potential impact on democratic elections. The research aimed to assess fears about AI’s persuasive capabilities given its access to vast information and user data.
