Generative Artificial Intelligence in Violence Risk Assessment: Emerging Technology and the Ethics of the Inevitable

Recent developments in artificial intelligence (AI) have stimulated considerable excitement and discussion regarding the potential impacts on people's lives and work. In particular, proposed and realized applications of generative AI have appeared across multiple industries and domains, including at the intersection of behavioral science and the law. This manuscript presents an ethical analysis of applications of generative AI to violence risk assessment, guided by the ethical principles of autonomy, beneficence and non-maleficence, and justice. The authors argue that generative AI, although capable of producing novel content, is nonetheless vulnerable to ethical problems, including through its exposure to biased training data. Issues such as limited transparency in decision making and the potential for the perpetuation and exacerbation of racial disparities are discussed. The authors recommend that professionals approach generative AI with due caution, as they would with any novel or emerging risk assessment approach, and suggest continued evaluation and research.