Conflict researchers face an unresolved dilemma: the underlying data are often unreliable. When it comes to covert relationships, killings, and illicit markets that organized violence entails, there are simply more incentives to alter information than to tell it straight. How confident can scholars be that on-the-ground events, rather than strategic or omitted information, drive research findings? Despite the evident need for accurate views into clandestine processes, existing work rarely applies systematic checks to verify the seeming “facts” of conflict. This article proposes a methodological toolkit to fill this gap. A first step develops systematic checks to report numerical credibility scores of source quality and corresponding error estimates. A second leverages data of varied strengths for distinct purposes: high-quality sources to triangulate facts and low-quality data to discern strategic images and mis/disinformation. The article tests these standards against major datasets and integrates the protocols into an interactive Data Evaluation Dashboard available for scholarly and policy use.

中文翻译：

---

了解我们不知道的：冲突研究中数据质量的根本问题——以及方法解决方案


冲突研究人员面临着一个未解决的困境：基础数据往往不可靠。当涉及到有组织暴力带来的秘密关系、杀戮和非法市场时，改变信息的动机比直截了当的说出来要多。学者们有多大的信心认为实地事件，而不是战略性或遗漏的信息，推动了研究结果？尽管显然需要对秘密过程的准确视图，但现有的工作很少应用系统性检查来验证冲突的看似“事实”。本文提出了一个方法论工具包来填补这一空白。第一步是开发系统检查，以报告来源质量的数字可信度分数和相应的误差估计。第二种利用不同强度的数据来实现不同的目的：高质量的来源来三角测量事实，低质量的数据来辨别战略图像和错误/虚假信息。本文针对主要数据集测试了这些标准，并将这些协议集成到一个交互式数据评估仪表板中，可用于学术和政策使用。