AI research has encountered significant ethical debates since its inception as a research discipline aiming to investigate,emulate,and augment human intelligence.The rapid progress in AI technology and the burgeoning proliferation of its technical applications has underscored the urgent and immediate necessity for the implementation of effective ethical governance in AI research.Despite significant efforts dedicated to ethical governance theory,there remains a lack of efficient practical methods due to the abstract nature of ethical theory.This study proposes AI ethical computation as a prospective approach to bridging the disconnect between ethical theory and ethical practice,providing a pathway to synchronize ethical principles with concrete applications.Based on its practical necessity and potential for development,the importance of ethical computation is clarified.Simultaneously,two paradigms of ethical computation methods for artificial intelligence are established based on the degree of ethical awareness and the autonomy of ethical decision-making as classification criteria.Through the abstraction of these paradigms,the study introduces three computation levels:ethical metrics,ethical inference,and ethical decision-making.Moreover,this study exemplifies ethical embedding and fair machine learning as instances to elucidate the characteristics and technical methods of the two research paradigms.The study concludes by presenting the construction of an ethical governance system and offering an outlook on the development of ethical computation.
NETL
NSTL
万方数据
