Abstract
Depending on the legal system, judges may have varying degrees of discretion in determining the type and extent of sentence that can be imposed for a particular offence. Nevertheless, it appears that even in systems traditionally considered discretionary, accepted patterns play a significant role in determining penalties, and judges utilize merely a limited spectrum of potential penalties in repetitive cases. Confirmation of the predictability of sentencing in certain categories of cases facilitates the possibility of automation. Utilising a computer program to assist judges in sentencing proposals based on input is feasible. This program can reflect the standard practice of sentencing penalties and punitive measures in a particular category of cases or rectify it, depending on the adopted sentencing policy. The objective of the article is to present findings from research that investigated whether a specific relation shapes the dimension of penalties and penal measures for cases concerning driving under the influence of alcohol in Poland, in the context of possible automation of the sentencing process. Another aim of this study is to provide an example of a straightforward mathematical recommendation model that tries to reflect both the discovered correlations in the data and the presumed intentions of legislators.