Abstract: Prescribing the fair analysis obligation of data processors in the form of legislation is a basic path to prevent algorithmic risks and promote algorithmic regulation.The setting of fair analysis obligation based on the concept of meta-regulation is not only a kind of procedural regulation,but also a kind of purposive regulation.This obligation requires the information processor to ensure that the analysis process is proper and that the analysis results are fair.Fair analysis obligation is broadly regulated in the legal system of personal information protection in China.According to the automated decision-making process,the fair analysis obligation can be classified into three categories:the obligation of care before analysis,the obligation to control during analysis,and the obligation to review after analysis.The normative orientation of the fair analysis obligation is to prevent algorithmic discrimination and treat individual citizens or data subjects fairly.Data processors that violate such obligations will bear administrative and civil liabilities,and the main way of accountability is to punish the phenomena of "big data price discrimination","information cocoons" and "social sorting".

Key words: automated decision-making, fair analysis obligation, algorithmic regulation