According to a new arXiv paper, the growing use of AI in high-stakes decision-making makes effective human-AI collaboration essential. The authors argue that successful collaboration requires humans to reflect on AI outputs rather than accept them automatically.

The paper's title, 'Think Once, Reflect Twice,' suggests that feedback shapes how humans reflect. The abstract frames reflection as a critical component of collaboration, and the title points to feedback as a mechanism for encouraging deeper consideration.

The abstract does not specify the methods or results, so the exact relationship between feedback and reflection is not detailed in the summary. Still, the paper highlights a design challenge for AI systems: how to prompt appropriate human reflection without undermining trust or efficiency.