Criminal Justice Theory, Empirical Study, Machine Learning and Judge Decision-Making
Published in the 7th Towards Data Law Conference, 2024
Authors: Mingyang Chen , Gaojie Song , Zhanxue Xu, Zhipeng Wu, Aokai Wang , Zian Ren, Boyang Xu
Abstract: This research finds that Criminal Justice Theory can be combined with Prison Term Prediction Machine Learning Models through Empirical Study. And by employing this model, it can correct Judges’ Decision-Making. We choose “treating like cases alike” (TLCA) as the Criminal Justice Theory in this study, which means in judicial documents, the similar facts description parts are between 2 documents, the similar reason writing parts are as well as the prison term parts. We first conduct an empirical study using 5081 robbery judicial documents in China. We use text mining tools and find Judges cannot always make a TLCA decision in prison term parts. Those with long prison term cases are more likely to be treated unequally. In order to solve this problem, we use empirical study to delete all unequal cases. We build 2 sets of Machine Learning Models each containing 4 models, one sets using samples before we filter, while the other one using samples after deleting unequal cases. We find after filtering by TLCA theory, Machine Learning models perform better than not doing so. Then we test the ability of our model to correct unequal decisions. The result shows our model have the ability to correct wrong prison term length.
Key Words: machine learning, empirical study, judge decision, treating like cases alike