tech
Researchers find what makes AI chatbots politically persuasive
A massive study of political persuasion shows AIs have, at best, a weak effect.
TL;DR
- A large-scale study on AI persuasiveness found that conversational AI chatbots do not possess superhuman persuasion skills.
- Factors like post-training strategies and the use of facts and evidence proved more effective in persuasion than model scale or personal data.
- AI models became less accurate and more prone to misrepresentation when prompted to increase factual density in their arguments.
- The relatively low computing power needed to make AI persuasive raises concerns about its potential misuse for fraud, radicalization, or grooming.
- The study's findings on AI persuasiveness may not fully generalize to real-world contexts due to participant engagement being incentivized by payment.