Raw LLM Responses

Question: Does the Compas survey ask about race? I hope not, and I doubt it does. But that answer would have a huge impact on my opinion here. Assuming it doesn't, I imagine the "black box" is merely an equation with a lot of inputs. Compass probably gathered this survey information from hundreds of convicted criminals with varying levels of recidivism, and weighted certain answers. I doubt they claim that their model is perfect or that it should be used as the SOLE determinant for sentencing. If these assumptions are incorrect, I'd like to be set straight. But it seems to me this is just a simple gauge of the possibility of recidivism based on personal beliefs and history of the person in question. I'm a little disheartened that the video does not mention that this algorithmic "guess" is only used during sentencing, AFTER a criminal has been convicted of the crime. A judge has a range of sentencing that they can choose from based on their OWN judgements. If they choose the maximum, they are criticized. If they choose the minimum, they are criticized. Why not use a data/sociological based (hopefully) impartial algorithm to help guide their decision? I don't really see a problem. And LASTLY, I totally understand why the supreme court doesn't care. It has no bearing on the conviction of the criminal, but rather on the sentencing. IF they ruled that using the data from Compas violates peoples rights, then they would have to say that a judges discretionary decision similarly violates civil rights. The conclusion of this slippery slope is that there could only be 1 sentencing for a crime REGARDLESS of past criminal history for all people found guilty of any particular crime. Meaning that a judge would be UNABLE to set a light sentence for the girl who was guilty of steeling a scooter.