Research on Tiered and Classified Governance of Algorithmic Decision-Making
As the core of artificial intelligence and other emerging information technologies,algorithms have been deeply embedded in business and public decision-making.However,the attributes and oper-ational characteristics of algorithmic automated decision-making,such as the coexistence of virtual and real decision-making entities,the iterative update of algorithmic technology,and the hiding and com-plexity of causal relationships,have made traditional decision-making governance architecture fre-quently ineffective.There still exist problems in the current algorithmic decision-making governance,which is mainly based on individual empowerment and due process control,including the lack of boundaries for algorithmic decision-making application,insufficient regulation,unclear public disclo-sure rules,and difficulties in identifying liability.Based on the superiority of tiered and classified al-gorithmic decision-making governance in terms of flexible regulation,risk identification and differenti-ated treatment,this paper constructs a tiered and classified model of algorithmic decision-making that takes into account the application scenarios and risk levels of algorithmic decision-making.The model divides algorithmic decision-making into four types of risk,and sets governance measures combining leniency with severity for different types of risk in view of entry thresholds,regulatory details,public disclosure levels,and scope of liability to create a scenario-based refined algorithmic decision-making governance mechanism in response to the complex architecture and diverse application scenarios of al-gorithmic decision-making.
algorithmic decision-makingindividual empowermentprocess controltiered and classi-fied governance
万方数据
维普
