Generative artificial intelligence holds enormous potential to revolutionize decision-making processes, from everyday to high-stake scenarios. However, as many decisions carry social implications, for AI to be a reliable assistant for decision-making it is crucial that it is able to capture the balance between self-interest and the interest of others. We investigate the ability of three of the most advanced chatbots to predict dictator game decisions across 78 experiments with human participants from 12 countries. We find that only GPT-4 (not Bard nor Bing) correctly captures qualitative behavioral patterns, identifying three major classes of behavior: self-interested, inequity-averse, and fully altruistic. Nonetheless, GPT-4 consistently overestimates other-regarding behavior, inflating the proportion of inequity-averse and fully altruistic participants. This bias has significant implications for AI developers and users.

翻译：生成式人工智能在从日常决策到高风险场景中具有变革决策过程的巨大潜力。然而，由于许多决策涉及社会影响，要使人工智能成为可靠的决策辅助工具，其必须能够捕捉自我利益与他人利益之间的平衡。我们研究了三个最先进聊天机器人预测独裁者博弈决策的能力，涉及来自12个国家78项实验的人类参与者。研究发现，仅有GPT-4（而非Bard或Bing）能正确捕捉定性行为模式，识别出三种主要行为类型：利己型、不平等厌恶型和完全利他型。尽管如此，GPT-4始终高估他人导向行为，夸大了不平等厌恶型和完全利他型参与者的比例。这一偏差对人工智能开发者和使用者具有重要启示。