Categories and legal boundaries of Algorithm-Based Sentencing

• In Event: Sentencing 7. Emerging Issues in Sentencing Research

Abstract

For some time now, there has been intense debate about the opportunities and limitations of algorithm-based sentencing. One hypothesis is that its use promises more equality in sentencing and, consequently, more relative sanction fairness. The technological progress of recent years, particularly the increased computing power and advancements in machine learning, has led to numerous practical implementations or attempts at implementation. Even more systems have been theoretically proposed. The legal evaluation of such systems cannot be generalized but must be appropriate to the diversity of the technical systems. The presentation proposes categorizing computer-assisted sentencing systems into 21 variants based on technical approach and influence on decision-making. This forms the basis for further considerations. Subsequently, a legal analysis will be conducted differentiated by categories, particularly considering the European Union's Artificial Intelligence Act. The current legal limits of each category will be examined. Based on this, it will be addressed whether further legislative action is needed.

Author

• Hauke Bock, University of Leipzig