인공지능 로봇의 형사책임 ―논의방향의 설정에 관한 몇 가지 발전적 제언―

Can thinking machines(Machina Sapiens) be subject to criminal law? If a robot erroneously identifies a man as a threat to its mission, and calculates that the most way to eliminate this threat is pushing him into an adjacent operating machine, who is to be held liable for this harm to a man. On this problem, there have been several conflicting arguments in legal academia. This paper tries to examine the validity of those arguments and show the desirable theoretical orientation for these arguments to follow and choose. As long as men used computers as mere tools, there was no difference between computers and hammer drill or telephones. When computers became considerably sophisticated, we used to say that computers ‘think’ for itself. The problem began when computers evolved from machines that were programmed to perform defined processes to thinking machines. For this reason, there has been significant controversy about the very essence of an AI entity. The paper argues that we cannot conclude that AI robot is criminally liable on the same level of criminal liability of humans, because from evolutionary viewpoint, the evolved psychological mechanism of humans is reluctant to impose the kind of criminal liability on AI entity. In addition, anthropocentric tribal prejudices against pseudo-humans such as corporate body or AI entity are also reluctant to regard those pseudo-humans as genuine humans. From this consideration, this paper shows that we can derive many implications for the conflicting arguments about the criminal liability of AI entity.