反恐预测算法“双黑箱”之对策研究——以美军反恐战争为例

Research on Countermeasures of the "Double Black Boxes" for Anti-Terrorism Prediction Algorithm——Taking America’s Anti-Terrorism as an Example

[目的/意义]智能反恐时代,反恐预测算法作为机器学习的应用算法,在预测恐怖活动组织和恐怖分子方面,效果显著。然而,应用中出现了“保密黑箱”叠加“技术黑箱”的“双黑箱”问题,对之研究具有重要的现实价值。[方法/过程]以美军反恐战争为研究样本,实证分析了造成“双黑箱”的原因,如数据偏差、算法歧视、过度依赖、透明不足和问责不力等;规范分析了美国政府破解“双黑箱”的政策和法律探索。[结果/结论]研究认为,确保可参与性、可诠释性和可问责性的透明化路径是解决反恐预测算法“双黑箱”的积极面向。

[Purpose/Significance]As anti-terrorism steps into the intelligent era, anti-terrorism prediction algorithm, as an application algorithm of machine learning, has a remarkable effect on predicting terrorist organizations and terrorists. However, the “confidentiality black box” and the “technology black box” problems, known as the“double black boxes” problems, which appear in intelligent anti-terrorism are the key of this research.[Method/Process]Taking the US military war on terror as a case, this research empirically analyzes the causes of “double black boxes”, such as data deviation, algorithm discrimination, over-reliance, insufficient transparency and ineffective supervision and accountability and normatively analyzes the policy and legal exploration of the US government.[Result/Conclusion]In conclusion, ways to solve the “double black boxes” problems of anti-terrorism prediction algorithm involve ensuring the transparency of interpretability, accountability and participation.