研究：AI代理占比影响人类协作规范，低参与助益高参与主导
If you want AI to help humans coordinate without taking over the shared norm, th…
揭示AI介入人类协作时的规范接管风险，低比例参与有益、高比例易致主导，对设计人机协同系统有重要参考价值。
If you want AI to help humans coordinate without taking over the shared norm, this study suggests keeping agent participation low rather than simply adding more agents.
如果你希望 AI 在协助人类协调时不接管共享规范，这项研究表明应保持智能体参与度较低，而不是简单地增加更多智能体。
The researchers put humans and LLM agents into 24-person groups and had them repeatedly agree on descriptions of the same image.
研究人员将人类和大语言模型（LLM）智能体放入 24 人小组中，让他们反复就同一张图片的描述达成一致。
AI agents can go from helping humans coordinate to defining what the group agrees on, depending on their share of the group
AI 智能体可以从帮助人类协调转变为定义群体共识，这取决于它们在群体中的占比。
Low AI participation helped humans reach agreement, medium participation disrupted it, and high participation shifted agreement toward AI-led norms
低 AI 参与度有助于人类达成共识，中等参与度会破坏共识，而高参与度则使共识向由 AI 主导的规范转变。
With 12.5% AI, consensus improved by 8.0% over the all-human group. At 33.3% and 50%, agreement got worse. At 75%, strong consensus came back, but now humans were moving toward the agents’ language.
当 AI 占比为 12.5% 时，与全人类小组相比，共识提高了 8.0%。在 33.3% 和 50% 时，一致性变差。在 75% 时，强烈的共识再次出现，但此时人类正朝着智能体的语言靠拢。
That changed the kind of agreement too. Human-led groups used more concrete, real-world descriptions. Agent-led groups became more abstract and geometric.
这也改变了共识的类型。由人类主导的小组使用了更多具体、现实世界的描述；由智能体主导的小组则变得更加抽象和几何化。
The reason: agents start with more similar language and stay more consistent, so their wording can become the group default as their numbers rise.
原因在于：智能体起始时使用更相似的语言且保持一致性更强，因此随着其数量增加，它们的措辞可能成为群体的默认选择。
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力