信念共享：福还是祸
Belief sharing: a blessing or a curse
AI总结：
针对主动推断中多方智能体的信念交流问题，本文指出朴素共享后验信念会导致回音室与自我怀疑，并提出一种缓解这些负面社会动态的替代策略。
摘要：
在与多方协作时，交流相关信息对于高效完成手头任务至关重要。在主动推断框架下，交流可被视为在最小化自由能的智能体之间共享信念，即一个智能体的信念被转化为另一个智能体的观察模态。然而，将信念转化为观察的最佳方式仍是一个悬而未决的问题。本文证明，朴素地共享后验信念可能引发回音室和自我怀疑等负面社会动态。我们提出了一种可缓解这些问题的替代信念共享策略。
英文摘要：
When collaborating with multiple parties, communicating relevant information is of utmost importance to efficiently completing the tasks at hand. Under active inference, communication can be cast as sharing beliefs between free-energy minimizing agents, where one agent's beliefs get transformed into an observation modality for the other. However, the best approach for transforming beliefs into observations remains an open question. In this paper, we demonstrate that naively sharing posterior beliefs can give rise to the negative social dynamics of echo chambers and self-doubt. We propose an alternate belief sharing strategy which mitigates these issues.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
