Intelligent machines are increasingly appearing in daily ethical scenarios, and there is an urgent need to consider the ethical module design of intelligent machines and give them moral considerations. In current research on machine ethics, the most discussed approaches are top-down and bottom-up approaches, both of which have their own rationality but have limitations in computability. Therefore, scholars have begun to study the design of hybrid paths, which involves how to better integrate the two approaches. This paper analyzes the problems encountered in the specific implementation of machine morality and proposes that both design and integration issues can adopt the method of transfer learning. It suggests improving the good ideas and suitable model frameworks applied in machine intelligence and then introducing them into the design and implementation of machine morality. Specifically, on one hand, the machine intelligence implementation method of “autonomous mental development” can be transferred to the framework design of “autonomous moral development”. On the other hand, considering the characteristics of moral issues themselves, factors such as emotion, sociality, semantic understanding, and the impact of consciousness on human moral decision-making should be taken into account and integrated into the machine morality module. Based on the above two considerations, this paper proposes a feasible modeling framework for machine ethics, providing a new design idea and practical reference for the realization of machine ethics.