Inspired by the increasing use of AI to augment humans, researchers have studied human-AI systems involving different tasks, systems, and populations. Despite such a large body of work, we lack a broad conceptual understanding of when combinations of humans and AI are better than either alone. Here, we addressed this question by conducting a meta-analysis of over 100 recent experimental studies reporting over 300 effect sizes. First, we found that, on average, human-AI combinations performed significantly worse than the best of humans or AI alone. Second, we found performance losses in tasks that involved making decisions and significantly greater gains in tasks that involved creating content. Finally, when humans outperformed AI alone, we found performance gains in the combination, but when the AI outperformed humans alone we found losses. These findings highlight the heterogeneity of the effects of human-AI collaboration and point to promising avenues for improving human-AI systems.

翻译：受人工智能日益广泛地用于增强人类能力的启发，研究者们已在不同任务、系统及人群背景下对人类-人工智能系统展开了研究。尽管已有大量研究成果，我们仍缺乏关于人机组合何时优于任何一方单独工作的整体性概念理解。为此，我们通过对100余项近期实验研究（报告了300多个效应量）进行元分析来探讨这一问题。首先，我们发现人机组合的平均表现显著差于人类或人工智能单独表现中的最优者。其次，我们发现涉及决策的任务存在性能损失，而在涉及内容创作的任务中则观察到显著更大的性能增益。最后，当人类单独表现优于人工智能时，人机组合呈现性能增益；但当人工智能单独表现优于人类时，人机组合反而出现性能损失。这些发现揭示了人机协同效应的异质性，并为改进人类-人工智能系统指出了具有前景的研究方向。