posted on Feb, 9 2013 @ 02:03 PM
If you think that the attributes and the methodology will not be modified to ensure equal outcomes across demograph groups, you're crazy. Lets all
take a look at the model, look at the suggested sentence and then the reality of the sentence and see if they even directionally correct.
You don't need sophisticated models. The attributes have been known for decades.
Born to a single mother with no education? 10 points
Over 20 with no work history? 10 points
No high school education? 10 points
Past run in(s) with the criminal justice system? 15 points
History of drug abuse? 10 points
Lives in a high crime neighborhood? 10 points
No outside interests, e.g. organized sports, music, art, volunteering, church, 5 points
Lets run the criminals through that model, which is highly predictive and see how that works.
The bottom-line is that the results will either be ignored or discounted either by simply over ruling the suggested sentence or manipulating the model
to get a demographly consistent outcome.