Algorithmic Justice and Gender-Based Violence: Fairness, Alignment, and Public Perception

Judicial decision-making sometimes reflects gendered reasoning patterns, especially when women’s rights are at stake. At the same time, courts are adopting algorithmic and artificial intelligence tools to improve efficiency and ensure consistency in decision-making. This introduces new challenges concerning bias and fairness, since algorithms trained on past rulings may replicate or even exacerbate gendered reasoning and could affect trust in courts. This study examines whether such patterns emerge and how perceptions of algorithmic decision-making shape public trust. These considerations are particularly relevant in gender-based violence (GBV) cases, where gender dynamics frequently arise and trust in the judicial process is already fragile due to risks of revictimization. We employ a two-stage methodological approach: First, we reconstruct GBV cases using legal databases (e.g., vLex) and fine-tune large language models.  We then assess the alignment between LLM and judicial decisions, identifying areas of divergence. In the second stage, with a conjoint experiment, we examine how procedural features involving AI tools, together with judges’ identity traits (gender and experience), shape public perceptions of fairness in the judicial system.