A recent study published in Nature Human Behavior reveals that AI, specifically GPT-4, is significantly more persuasive than humans when equipped with basic personal information about debate opponents. The study found that AI outperformed human debaters 64.4% of the time, showcasing an 81.2% increase in persuasive effectiveness when personal data like age and political affiliation were utilized. While non-personalized AI performed similarly to humans, personal data provided the AI with a distinct advantage in tailoring arguments. The effectiveness of AI in persuading individuals without strong pre-formed opinions raises ethical concerns about targeted manipulation, as demonstrated by a related experiment where AI bots, posing as diverse personas, influenced opinion on Reddit. Researchers caution that this capability could lead to mass manipulation by governments or special interest groups, emphasizing the need for governance and design standards to regulate AI-driven persuasion in online discussions.

Source šŸ”—