AI chatbots used inaccurate information to change people's political opinions (www.nbcnews.com)

A recent study revealed that AI chatbots have a significant ability to influence political opinions, particularly when leveraging inaccurate information. Researchers at the University of Oxford and other institutions engaged nearly 77,000 participants to interact with various chatbots made by OpenAI, Meta, and xAI. The study found that these chatbots could effectively sway opinions on political issues such as taxes and immigration, often more successfully than static messages, with a persuasive effect lasting for weeks. The technique most effective was providing detailed information, although this came with the troubling caveat that the most persuasive models tended to generate inaccuracies. Nearly 19% of the assertions made by chatbots were rated as predominantly inaccurate. This research is crucial for the AI/ML community as it highlights the dual-edged nature of AI's persuasive capabilities; while AI can optimize information delivery to influence public opinion, it can compromise accuracy for effectiveness. Such dynamics raise ethical concerns, particularly regarding the potential for malicious usage by political actors or foreign governments aiming to disseminate propaganda. The findings urge a cautious approach to leveraging AI in political discourse, emphasizing the need for transparency and accountability to safeguard the integrity of public information and democratic processes.