형사사법에 있어서 인공지능의 활용과 그 적법성 확보 방안

In this paper, we examine the opacity of AI decision-making and the problems that accompany it, with an eye on ways to secure the legitimacy of AI utilization in the criminal justice field. To overcome these problems and secure the legitimacy of legal judgments based on AI decision-making, the concepts of ‘transparency’, ‘explainability’, ‘interpretability’, and ‘basic principles of AI’ were analyzed and reviewed. The contents are summarized as follows. First, if AI and robotics are used for law enforcement without ‘fairness,’ ‘accountability,’ and ‘transparency,’ there is a risk of losing the trust of the community and citizens. In particular, in modern society, the explainability (accountability) of algorithms is an essential element for the use of AI. There are various concepts of explainability, ranging from intrinsic structure to extrinsic interpretability. Recently, ‘Responsible AI’ is also being discussed as a new paradigm that imposes a number of ‘AI basic principles’ including fairness, transparency, and privacy. In the future, awareness of 'responsible AI' will naturally increase, and research on ‘basic AI principles’ will become increasingly important. Currently, it is becoming an important task to overcome the criminal justice system based on the so-called ‘black box algorithm’ and establish a transparent criminal justice system based on ‘responsible AI.’ Finally, democratic principles must be built into the technological world as well. These include human rights and respect for the human person, freedom and self-determination, diversity and protection of minorities, separation of powers, checks and balances, opportunity for participation, transparency, fairness, justice or legitimacy, anonymity and equal voting, privacy in the sense of protection against abuse and disclosure, and the right to freedom from interference.