Trust mechanisms diverge between centralized and decentralized exchanges, representing distinct sociotechnical governance paradigms. However, quantifying trust dynamics and their redistribution between these architectures remains empirically challenging, limiting understanding of how institutional shocks affect market behavior. The FTX collapse offers a natural experiment to bridge this gap. Through an interdisciplinary approach combining causal inference and computational text analysis, we find significant price declines and capital reallocation from centralized to decentralized exchanges following the event. While sentiment metrics showed no sharp discontinuities, topic modeling and network analysis of Discord communities reveal that seasonal holiday discourse obscured underlying trust concerns in centralized exchange forums. These findings underscore the fragility of institutional trust architectures and demonstrate how mixed methods can illuminate behavioral patterns during systemic crises, offering insights for exchange risk management and regulatory assessment.

翻译：中心化交易所与去中心化交易所的信任机制存在差异，代表了截然不同的社会技术治理范式。然而，量化这些架构之间的信任动态及其再分配在实证研究中仍具挑战性，这限制了我们对机构冲击如何影响市场行为的理解。FTX的崩盘为弥合这一差距提供了自然实验。通过结合因果推断与计算文本分析的跨学科方法，我们发现事件发生后，加密货币价格显著下跌，资本从中心化交易所重新流向去中心化交易所。尽管情绪指标未出现急剧不连续性，但Discord社区的主题建模和网络分析揭示，季节性节日话语掩盖了中心化交易所论坛中潜在的信任担忧。这些发现凸显了机构信任架构的脆弱性，并展示了混合方法如何在系统性危机中揭示行为模式，为交易所风险管理和监管评估提供洞见。