The study by Vaccaro, Almaatouq, and Malone explores the effectiveness of human-AI systems in various tasks, systems, and populations. Despite the growing use of AI to augment human capabilities, there is a lack of a broad conceptual understanding of when human-AI combinations are better than either humans or AI alone. The researchers conducted a meta-analysis of over 100 recent experimental studies reporting more than 300 effect sizes to address this question. Key findings include: 1. On average, human-AI combinations performed significantly worse than the best of humans or AI alone. 2. Performance losses occurred in tasks involving decision-making, while significant gains were observed in tasks involving content creation. 3. When humans outperformed AI alone, performance gains were found in the combination, but when AI outperformed humans alone, performance losses occurred. The study highlights the heterogeneity of human-AI collaboration and identifies factors that influence synergy, such as task type, relative performance of humans and AI, and experimental design. The results suggest that while human-AI systems often perform worse on average, they can still augment human performance in specific contexts, particularly in content creation tasks. The authors also provide recommendations for future research, including developing generative AI for creation tasks, innovative processes for human-AI collaboration, more robust evaluation criteria, and commensurability criteria to facilitate systematic comparisons across studies.The study by Vaccaro, Almaatouq, and Malone explores the effectiveness of human-AI systems in various tasks, systems, and populations. Despite the growing use of AI to augment human capabilities, there is a lack of a broad conceptual understanding of when human-AI combinations are better than either humans or AI alone. The researchers conducted a meta-analysis of over 100 recent experimental studies reporting more than 300 effect sizes to address this question. Key findings include: 1. On average, human-AI combinations performed significantly worse than the best of humans or AI alone. 2. Performance losses occurred in tasks involving decision-making, while significant gains were observed in tasks involving content creation. 3. When humans outperformed AI alone, performance gains were found in the combination, but when AI outperformed humans alone, performance losses occurred. The study highlights the heterogeneity of human-AI collaboration and identifies factors that influence synergy, such as task type, relative performance of humans and AI, and experimental design. The results suggest that while human-AI systems often perform worse on average, they can still augment human performance in specific contexts, particularly in content creation tasks. The authors also provide recommendations for future research, including developing generative AI for creation tasks, innovative processes for human-AI collaboration, more robust evaluation criteria, and commensurability criteria to facilitate systematic comparisons across studies.