On The Risk Governance Dimension Algorithm Recommended Behavior Differentiation Regulation

Algorithm recommendation behavior is the specific application of the limited autonomous tool of recommendation algorithm.With the continuous development of science and technology,algorithm recommendation behavior not only brings a lot of convenience to social life,but also produces many risks such as algorithm discrimination,algorithm exploitation,algorithm black box and algorithm rule because of the particularity of its design purpose,data base,internal structure and decision-making mode.In the traditional sense,the algorithm regulation usually takes the data as the starting point and the scene as the regulation origin to carry out the linear hierarchical regulation with the preemptive regulation as the main.This cannot timely respond to the multiple sources of behavioral risks recommended by the algorithm and the progressive pattern of constant development,nor can the extensive differentiation mode reflect the reasonable regulation of high-risk algorithms and the coordination of the overall incentive development of the industry.This paper makes an in-depth study of algorithm recommendation behavior,and finds that algorithm recommendation behavior not only produces existing risk types caused by common tool abuse,but also produces newer and wider range of potential risk patterns due to its complex,intelligent and constantly developing characteristics.Therefore,based on the characteristics of algorithm recommendation behavior,the four key structures of data,field,architecture and AI are classified and analyzed,and summarized into two levels: application and technology.At the same time,the risk governance logic is introduced,and the existing risks and potential risks are typed and matched with the structure of the algorithm’s recommended behavior,so as to accurately identify the source of risks.After a comparative analysis of the current situation of domestic and foreign regulations,this paper holds that: in order to reflect the concept of accurate and efficient governance,on the basis of summarizing the existing regulatory experience of risk governance guidance classification and classification outside the region,risk elements and their sub-elements should be summarized according to risk categories and risk sources for classification.Then,based on the types and sizes of risks,the traditional linear classification system is abandoned,and algorithms of different types of risks are separately classified in application and technology,so as to promote the development of algorithm recommendation technology and ensure its orderly development without leading to the alienation of the original order function.In order to make the algorithm recommendation behavior run reasonably and develop sustainably in the regulation,it is necessary to conduct targeted normative design for algorithms with different risk levels in the whole stage of function design,operation deployment and damage occurrence based on the classification system,so as to realize targeted differentiated regulation.At the same time,we should clarify and improve the overall rules for the operation of the recommendation algorithm at the level of legislation and government,market and industry,citizens and society,and guide the algorithm recommendation activity to develop to a better direction under the spirit of collaborative governance and regulation.