Abstract: Driving behavior abides by two-tier morality:The first tier is the morality of drivers,and the high-order tier is the morality of ethical experts.The “first-tier morality” is applied to daily driving behaviors,while the “high-order morality” emerges in the “trolley problem” thought experiment.Embedding the “first-tier moral framework” into autonomous driving systems in the form of algorithms to enable such systems to act as driving experts constitutes the current “responsibility-based action ethics” for autonomous driving to meet the travel demands for safety,efficiency,and comfort.This framework is based on the first-tier nature of specialized artificial intelligence.The introduction of “high-order morality” into ethics to address responsibility dilemmas has sparked discussions on what kind of moral algorithms should be embedded in autonomous driving,which is rooted in the high-order nature of general artificial intelligence.From the perspective of functional reasoning,we can compare the “two-tier moral framework” and its application in autonomous driving.We should neither “totally oppose” nor “overestimate” the function of high-order moral algorithms.Instead,we should position them within the framework of “preparatory responsibility-based action ethics”.Starting from the normative demand for preventing AI ethical risks,it is necessary for humans to prepare high-order moral algorithms for autonomous driving.It serves as an indispensable “preparatory action ethics solution” configured at the general level for mitigating ethical risks in autonomous driving,with the function of “supplementing responsibility gaps” (i.e.,addressing “responsibility deficiencies”).

Key words: autonomous driving, high-order moral algorithms, trolley problem, preparatory responsibility action ethics