Coding Anti-Discrimination Jurisprudence: A Hybrid Computational Model of the Arlington Heights Test

The aim of this study is to understand the judicial reasoning process of anti-discrimination jurisprudence by utilizing a hybrid computational model. The advancement of hybrid computational legal studies that combine “law-as-code” and “law-as-data” approaches have led to promising techniques for tackling complex legal reasoning tasks as multifactor judicial reasoning standards. Following this hybrid model, this study conducts a statistical and multilayer perceptron (MLP) analysis of the judicial reasoning process of the multifactor Arlington Heights discriminatory purpose test based on an original hand-coded dataset of discrimination cases. The results of the study show that “sequence of events” predominates the other factors for predicting the outcome of discrimination cases, with “statistical impact” and “historical background” showing particularly weak effect on the outcome. Results show that completely removing the factor of disparate statistical impact (Impact) actually improves confidence in predicting discrimination. The conclusion of the study suggests that courts fail to give either disparate statistical impact or historical record of discrimination any meaningful cumulative import in determining discriminatory purpose. I conclude by observing the benefits of hybrid computational legal studies on interrogating more normative values in the law like transparency, fairness, and procedural justice.