- Abstract: Legal fairness is one of the most important principles pursued by modern legal systems. Unfortunately, unfairness may be inevitably introduced in real-world cases due to both objective and subjective uncertainty, such as ambiguity in the law or practical bias in judgments. Existing works for fairness analysis mainly rely on labor-intensive element annotation for cases, which suffer from limited generalization ability. To address this issue, we propose to utilize large-scale textual data to perform quantitative legal fairness analysis via our Causal-based Legal Fairness Measuring Framework (CaLF). To verify its effectiveness, we construct a legal-fairness dataset, and experimental results show that CaLF can accurately characterize the unfairness. Further, we adopt CaLF on a large-scale real-world dataset. Based on our settings and within our dataset, we have several interesting experimental observations from the perspective of gender, age, and region.