算法透明度的理论反思与制度建构

Theoretical Reflection and Institutional Construction of Algorithmic Transparency

算法透明度在整个算法治理实践中蕴藏着丰富的固有价值和工具性价值。当前基于信息的算法透明度模式在理论渊源上源自传统基于信息的透明度理论,该理论虽发挥了一定作用,但因忽视算法决策的复杂性和动态性,难以满足利益相关者的多元化透明需求,忽视了普通公众的算法理解力而面临着严重的实践困境。基于此,在理论层面上,宜对算法透明度进行合理的理论更新,倡导算法透明度的“关系”转向,由传统静态的基于信息的算法透明度模式迈向动态的基于关系的算法透明度模式。在具体制度建构上,宜以关系透明度理论为指导,遵循“透明度设计”理念,基于算法透明度的个体和系统双重维度优化、完善算法透明度的基础性制度,并基于双阶透明度探索建构具有实践理性的算法透明具体机制,以增强算法透明度制度的回应性和有效性,切实发挥算法透明度在算法问责和算法治理中的功能。

Algorithmic transparency has rich inherent value and instrumental value in the whole algorithmic governance practice.The current information-based algorithmic transparency model is rooted in the traditional information-based transparency theory in its theoretical origin.Although it plays a certain role,it is difficult to meet the diversified transparency needs of stakeholders due to the neglect of the complexity and dynamic characteristics of algorithmic decision-making,and the neglect of the general public’s understanding of the algorithm,and faces serious practical difficulties.Based on this,on the theoretical level,it is advisable to update the theory of algorithmic transparency reasonably,advocate the“relational turning”of algorithmic transparency,from the traditional static information-based algorithmic transparency model to the relation-based algorithmic transparency model.In terms of specific institutional construction,it is advisable to follow the concept of“transparency design”under the guidance of the theory of relational transparency,build a basic system of algorithmic transparency based on the individual and system dimensions of algorithmic transparency,and explore a practical concrete mechanism of algorithmic transparency based on the two order of transparency,so as to enhance the responsiveness and effectiveness of the algorithmic transparency system and effectively exert the institutional value of algorithmic transparency in algorithmic accountability and governance.