如何实现“黑箱”下的算法治理?——平台推荐算法监管的测量实验与策略探索

Algorithms Governance in the“Black-Box”:Experimentation and Exploration of Platform Recommendation Algorithms Supervision

近年来,平台推荐算法的快速发展和广泛应用深刻改变了互联网信息内容的供给方式,同时也引发了一系列算法风险与现实问题。平台推荐算法治理与监管成为政府规范算法应用的重点内容之一,无论是算法备案制度的探索,还是实现算法透明化的设想,均面临着一定的困难和挑战。在不打开算法“黑箱”的前提下,平台推荐算法规制与监管是否有可行之道?论文基于机器行为学思想,采用实验方法,以用户视角对推荐结果进行跟踪记录,通过实验数据的对比分析,验证了平台推荐算法结果差异的可测性。基于实验结果,论文提出了通过对不同平台推荐结果进行大规模数据测试和检验,测量平台推荐算法运行逻辑与推荐效果,从而实现“黑箱”下有效的逆向监管,以期丰富未来算法治理的选择。

In recent years,the rapid development and widespread application of platform recommendation algorithms have changed the supply mode of Internet information content,while causing a series of risks and practical challenges.The governance and supervision of platform recommendation algorithms have become key aspects of government regulation of algorithm applications.The implementation of algorithm filing system and algorithm transparency are faced with difficulties and challenges.Is there a feasible way for the regulation and supervision of platform recommendation algorithms without opening the algorithm“black-box”?Based on the machine behavior and the experimental methods,the study tracked and recorded recommendation results from the user perspective.Through comparative analysis of experimental data,the measurability of the differences in platform recommendation algorithms results was verified.Based on the experimental results,this work proposes a set of measure schemes of the operational logic and recommendation effectiveness of platform recommendation algorithms based on a large-scale data testing and verification on recommendation results from different platforms for achieving substantive supervision in the“black-box”and enriching approach choices of algorithms governance in future.