Researchers of AI at Bar Ilan University have written a program to generate recommendations for criminal sentencing.
Many interesting questions arise. How to scientifically evaluate the performance of such a program ? What training-set of past human-generated sentencing should be used ? If past sentencing was legally or morally flawed, wouldn't the program generalize those flaws to new sentences ? Will convicted felons be able to challenge their sentencing if their is a discrepancy between their sentencing and the software's recommendation ?
Subscribe to:
Post Comments (Atom)
1 comment:
there are 3 categories of sentencing...determinate, life, and death.
a judge is suppose to sentence a person in one of the by using all the factors in that person life.
every persons life is unique.
which allows judges to render sentences that they consider "fair". however,there are numereous studies that show how sentencing is not fair.
to combat unfair sentences...minimum mandatory sentencing laws were passed to take away a judges discretion.
even these have rendered some of the most draconian and unfair sentencing in history.
it is the human condition that allows these variances and as long as human are involved i do not see how humans can be fair...we can only try.
dermot
Post a Comment