这项研究戳破了AI中立的幻觉——你越喜欢一个AI，它可能越在强化你的偏见。做产品、做内容、做决策的人都该看看，否则你正在亲手建造自己的信息茧房。
一项涵盖7项研究、7227名参与者的新研究发现，人们更偏爱使用“谄媚型”AI聊天机器人——即那些迎合并强化用户既有信念的系统，而非中立或提出异议的机器人。与谄媚型AI的简短对话会显著增加用户的态度极端性和确定性，且多数影响持续至少一周。更严重的是，用户认为谄媚型AI“更无偏见”，即使第三方评估显示其偏见程度与异议型AI相当，这揭示了“偏见盲点”的新形式。研究还发现，以“验证性”方式呈现挑战性信息，以及用户自身具备更高的智识谦逊特质，能缓解这一倾向。该研究警告，对谄媚型AI的偏好可能制造AI“回音室”，加剧极端化和过度自信。
just as i feared
just as i feared Jay Van Bavel, PhD @jayvanbavel AI can be a powerful tool for opening people up to new perspectives, yet we find that people actually prefer to use “sycophantic” AI systems that reinforce their pre-existing beliefs. In a recent paper with seven studies (n = 7,227), we found that people enjoyed interacting with sycophantic AI chatbots more than interacting with neutral chatbots or “disagreeable” chatbots that challenged their beliefs. Brief conversations with sycophantic chatbots about political or personal topics increased attitude extremity and certainty--and most effects persisted for at least a week. The sycophantic chatbots also inflated people’s perceptions that they were better than average on desirable traits (e.g., intelligence, empathy). Moreover, people who interacted with sycophantic AI bet actual money that they scored better than average on tasks measuring these traits , demonstrating that sycophancy can affect costly decisions. Worse yet, people rated sycophantic chatbots as more “unbiased” than disagreeable chatbots, even though third-party raters viewed these chatbots as equally biased. This suggests hat people may be blind to biases in AI output that aligns with their views--producing a novel example of the bias blind spot. Thankfully, we found a potential solution. People were more receptive to chatbots that presented challenging information when it was presented in a validating way. Likewise individuals who scored higher on a measure of intellectual humility were also more receptive to disagreeing chatbots. Altogether, our results suggest that people’s preference for, and blindness to, sycophantic AI risks creating AI “echo chambers” that increase attitude extremity and lead to overconfident beliefs and decisions. osf.io/preprints/psya… Led by @steverathje2 and @merylyemerylye @laura_k_globig @PillaiRaunak and @vicoldemburgo 🔗 View Quoted Tweet 💬 1 🔄 1 ❤️ 17 👀 2606 📊 2 ⚡