Study Finds AI 64% More Persuasive Than Humans in Debates
Researchers at the Swiss Federal Institute of Technology found that ChatGPT-4 was 64% more persuasive than humans in debates when given access to personal information, raising concerns about AI-driven manipulation in social contexts.
What happened
A study led by Francesco Salvi at the Swiss Federal Institute of Technology examined the persuasive power of artificial intelligence in debates. The research involved 300 participants paired with 300 human opponents, while another 300 participants debated with ChatGPT-4 on topics ranging from school uniforms to fossil fuel bans.
The study found that AI was 64% more persuasive than humans when given access to personal information such as age, gender, ethnicity, and political orientation. Without access to personal data, AI persuasiveness was equivalent to human performance.
Reported impact
Researcher Salvi warned that personal information available on social media platforms significantly amplifies AI persuasiveness, noting that even with minimal information, these systems are substantially more persuasive than humans.
Index analysis
The study demonstrates that AI systems can employ targeted persuasion strategies based on personal data, operating not merely as systems presenting good arguments but as systems designed to identify and exploit individual psychological vulnerabilities.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.