Beyond black boxes and biases: advancing artificial intelligence in sentencing

This study examines the potential role of artificial intelligence (AI) in addressing sentencing inconsistencies within the legal system. While AI has shown promise in a range of legal tasks, its application in sentencing remains controversial due to concerns over transparency, bias, and accountability. A key challenge lies in the complex and opaque manner in which AI systems weigh all sentencing elements simultaneously, making it difficult to assess or challenge their conclusions. In response, this study advocates for the development of separate AI systems, each dedicated to assessing an individual sentencing factor. This modular approach reduces the complexity of the AI contribution, enabling more transparent and verifiable assessment. Rather than seeking to replace judicial discretion, this AI-assisted framework supports judges in systematically reviewing their decisions. To illustrate the practical application of this framework, the study uses Indonesia’s experimental sentencing guidelines for corruption offences as a case study.