Han-Wei Ho (Center for Empirical Legal Studies), Keng Wei Fan (National Dong Hwa University), and Kuan-ting Chen (Cornell University Law School; National Chengchi University) have posted How to Measure “Treat Like Cases Alike” in Sentencing: A Manifold-Learning-Based Analysis of Capital Cases in Taiwan on SSRN. Here is the abstract:
Preventing arbitrary death sentences is a fundamental human rights goal and a key objective for judiciaries worldwide. However, the rigorous examination of potential arbitrariness presents significant methodological challenges. While common empirical approaches frequently revolve around racial disparity via regressions—a focus well-accepted in the United States due to historical injustices and the role of sentencing guidelines—no obvious demographic bias has been identified in the existing literature regarding Taiwan’s death penalty sentencing. In such a context, there is an ongoing need to evaluate potential arbitrariness holistically and to supplement traditional models that may not capture the full complexity of judicial discretion.
This study proposes a methodology to identify potential arbitrariness by shifting the analytical focus to judicial “noise”—defined as inconsistency that undermines the principle of “treating like cases alike”. Based on a hand-collected dataset of 336 Taiwanese death penalty cases (2013–2024) featuring 149 sentencing factors, UMAP is employed to characterize case similarity. As a manifold-learning algorithm, UMAP projects high-dimensional space into a two-dimensional representation by retaining the topological relationships and proximity between observations from the original high-dimensional space. This approach facilitates a rigorous evaluation of case similarity.
The results reveal instances of spatial inconsistency: several death penalty cases are observed spatially adjacent to cases involving “non-most serious crimes” or even fixed-term imprisonment. This immediate spatial proximity—representing similar case characteristics in the high-dimensional space—between different outcomes provides prima facie reason to further examine potential arbitrariness. By demonstrating UMAP’s utility, this study offers an additional empirical direction for identifying out-of-balance cases in jurisdictions where judicial consistency constitutes a primary concern for the rule of law.
To receive new posts from Legal Theory Blog by email, get a free subscription to Legal Theory Stack.
Lawrence Solum