Exploring the Causal Relationship Between Forgiveness and Sentencing Outcomes in Dangerous Driving Cases

Abstract

In the theoretical framework of criminal law in our country, the study of crime is primarily focused on the perpetrator, and criminal legislation is predominantly concerned with defining crimes from the perspective of the perpetrator. This has resulted in a relatively weak understanding of the behavior of victims in criminal studies. However, the crime of dangerous driving is a typical minor offense in our criminal punishment system and a common traffic safety crime, making the issue of sentencing in such cases a matter of concern. This article aims to explore the causal effect of whether the victim forgives the offender on the severity of sentencing in dangerous driving cases. Methods: Firstly, based on various statutory and discretionary sentencing factors, features are extracted from judicial documents using a combination of regular expressions and annotation. After determining the significance of these features, variable selection is performed. A sentencing prediction model is constructed using the Stacking method on the selected variables. Then, causal effects of forgiveness on the severity of sentencing are estimated using conditional outcome model, grouped conditional outcome model, inverse probability weighting, and doubly robust method. Finally, bootstrap sampling is used to provide interval estimates of causal effects under different models. Results: Our research results indicate that, on average, defendants who receive forgiveness from the victim are 32.7% more likely to receive lighter penalties. Quantifying the impact of forgiveness on sentencing outcomes through causal inference methods helps clarify the role of forgiveness in the sentencing system of criminal cases in our country. It also addresses the insufficient research on the post-crime factor of victim attitude in criminal and punishment theories, providing further data references for the standardization of judicial practices in our country.

This work was partly supported by National Key R &D Program of China under Grant 2020YFC0832400. This work was partly supported by Key R &D Program of Sichuan Province under Grant 2021YFS0397.