算法透明的法律实现与限定——基于商业秘密保护的视角

Legal Realization and Limitation of Algorithmic Transparency——Based on the Perspective of Trade Secret Protection

伴随着算法应用中个人自治价值不断被消解、社会秩序价值受到侵害以及传统治理框架频繁失灵,尤其是以ChatGPT为代表的基于统计学习神经网络的典型黑箱模型对现有问责机制带来了极大挑战,算法透明引发各国立法者高度关注。然而,提升算法透明度与保护商业秘密的规范目标之间存在紧张关系。算法自动化决策背后存在多元价值诉求的叠加,这体现为不可避免地将个人利益、公共利益、经济效率、公平正义等价值相互交织在一起,使得实现算法透明与保护算法商业秘密二者之间的取舍成为难题。主流利益平衡理论仅从利益位序去协调二者之间冲突的思路,具有局限性,应当在“调和论”视角下以“促进算法经济的高质量发展”为目标相互协调:一方面,应选择适当的算法透明工具,构筑算法解释的体系化框架,同时辅以多元算法监督机制;另一方面,应合理保护算法商业秘密,完善商业秘密的权利限制制度。

With the continuous elimination of the value of personal autonomy in the application of algorithms,the violation of the value of social order,and the frequent failure of traditional governance frameworks,the typical black-box model based on statistical learning neural networks represented by ChatGPT has brought great challenges to the existing accountability mechanism.As a big challenge,algorithm transparency has aroused great attention from lawmakers in various countries.However,there is a large tension between the normative goal of improving algorithmic transparency and protecting trade secrets.The superposition of multiple value appeals behind the algorithmic automated decision-making inevitably interweaves personal interests,public interests,economic efficiency,fairness and justice and other values,making the trade-off between realizing algorithm transparency and protecting algorithm business secrets a difficult problem.As a solution to the conflict of interests,the theory of interest balance is limited in the idea of coordinating the conflicts between the two only from the order of interests.It should be based on the perspective of“reconciliation theory”with the goal of“promoting the high-quality development of the algorithmic economy”to coordinate:on the one hand,select appropriate algorithm transparency tools,build a systematic framework for algorithm interpretation,and supplement it with a multi-algorithm supervision mechanism;on the other hand,seek reasonable protection of algorithmic trade secrets,and improve the system of restricting rights of trade secrets.