Ethical Judgment using Large Language Model

Robot ethics, also known as roboethics, is becoming increasingly necessary, especially in domestic use and autonomous driving. In this paper, we investigate the ethical decision-making process of ChatGPT, focusing on the keyword “Ethics.” We examine whether ChatGPT behaves in a cooperative, uncooperative, or unrelated toward humans. For ethical judgment, we employ three frameworks: utilitarianism, deontology, and virtue ethics. The results reveal similar judgment tendencies across all ethical frameworks. In the experiment, we prepared 50 sentences considered ethically correct, 50 neutral sentences, and 50 sentences considered ethically wrong from the Internet and conducted an ethical evaluation using ChatGPT. Our experiments show that using ChatGPT4 improves ethical judgment; however, caution should be exercised when using ChatGPT at this time.