Abstract

This paper conceptualizes the shortcomings of the manual jury selection process; a feature we perceive as a legal deficit with detrimental consequences to the judicial systems across nations. It introduces a pioneering operationalization: The PereiraJuryNet. We contribute by outlining the jury selection processes and a subsequent assessment of the problems associated with the manual selection of jurors by the prosecution and defense. Mindful of the inefficiencies of the manual jury selection system we advocate the use of machine learning algorithms in jury selection devoid of human bias and prejudice. We developed the PereiraJuryNet using Random Forest, SVM, Adaboost and Naive-Bayes machine learning models. Our proposed model recognizes key traits commonly considered during jury selection and voir dire such as political preference and awareness, conflicts of interest, demographic traits, experience and knowledge etc. and cross- references these with the nature and facts of diverse cases thereby being able to generate the 12 jurors best suited for any given case. Further, we illustrate the need, mechanics, data science and machine learning pipeline of the PereiraJuryNet model and provide suggestions for future work.

Keywords