Countries around the world are increasingly investing in artificial intelligence (AI) to automate military tasks that traditionally required human involvement. Despite growing interest in AI-enabled systems, relatively little research explores whether and how AI affects military decision-making. Yet, national security practitioners may perceive the judgments of and actions taken by algorithms differently than those of humans. This variation may subsequently affect decisions on the use of force. Using two original survey experiments fielded on a sample of US national security experts, we find that AI use by both friendly and rival forces affects decision-making during interstate crises. National security experts are less likely to take military action when AI is used to analyze intelligence than when humans conduct the analysis. Experts also viewed an accident involving a rival's AI-enabled weapon that kills American troops as more deserving of retaliation than an accident involving only human operators, suggesting that national security practitioners are less forgiving of errant AI systems than of similarly erring humans. Our findings suggest emerging technologies such as AI can affect decisionmakers’ perceptions in ways that shape political outcomes. Even in a world of algorithms, human decisions will still have important consequences for international security.

中文翻译：

---

算法与影响人工智能与危机决策

世界各国越来越多地投资于人工智能 (AI)，以实现传统上需要人类参与的军事任务自动化。尽管人们对支持人工智能的系统越来越感兴趣，但关于人工智能是否以及如何影响军事决策的研究相对较少。然而，国家安全从业者对算法的判断和采取的行动可能与人类不同。这种变化可能随后会影响有关使用武力的决定。通过对美国国家安全专家样本进行的两项原始调查实验，我们发现友军和敌对部队对人工智能的使用都会影响州际危机期间的决策。与人类进行分析相比，使用人工智能分析情报时，国家安全专家不太可能采取军事行动。专家们还认为，与仅涉及人类操作员的事故相比，涉及竞争对手的人工智能武器导致美军死亡的事故更值得报复，这表明国家安全从业人员对错误的人工智能系统的宽容度不如对类似错误的人类的宽容。我们的研究结果表明，人工智能等新兴技术可以影响决策者的看法，从而影响政治结果。即使在一个算法世界中，人类的决定仍然会对国际安全产生重要影响。我们的研究结果表明，人工智能等新兴技术可以影响决策者的看法，从而影响政治结果。即使在一个算法世界中，人类的决定仍然会对国际安全产生重要影响。我们的研究结果表明，人工智能等新兴技术可以影响决策者的看法，从而影响政治结果。即使在一个算法世界中，人类的决定仍然会对国际安全产生重要影响。