15 September 2016

Is Math Racist?

Via Gruntled Center, I found this article which leaves me wondering if the author has ever actually been involved in a criminal sentencing. She seems to assume some sort of godlike powers for both those putting together statistical frameworks for sentencing and those filling out the guideline sheets that are handed down.

Here's the statement which piqued my interest:
"[S]ome states started using recidivism models to guide sentencing. These take into account things like prior convictions, where you live, drug and alcohol use, previous police encounters, and criminal records of friends and family. "
(1) Prior convictions - Yes, they are used.

(2) Where you live - Really? In what world is this ever considered statistically? Virginia certainly does not (see Virginia guidelines for drug crimes as an example) and I doubt many other States have the resources to develop and keep up the kind of state-wide models that would be required. I haven't done any federal work in over ten years, but a look through their guidelines manual doesn't show anything along those lines either (see particularly chapters three and four found here). Sure, a specific judge may have enough knowledge to know what an area of his city/county is like and take it into account, but I want to see proof before I believe it is being put to significant statistical use.

(3) Drug and alcohol use - While this could be a valid consideration in sentencing (for reasons that could both be helpful or harmful to a defendant), I've never seen it considered in the statistical portion of sentencing guidelines. There is usually a section in the social history accompanying the guidelines that addresses this, but it's not part of the statistics.

(4) Previous police encounters - I don't know about other States, but introducing unadjudicated crimes and police encounters during sentencing is a big no-no where I've practiced in Virginia and if I tried to use something like that (without the defense attorney bringing it up first) I would get a dressing down by the judge. In the federal system things work differently and unadjudicated offenses tied in with the one you have been convicted of can be added in. But even there "previous police encounters" alone would not be statistically added in.

(5) Criminal records of friends and family - Again, in what world is this ever considered statistically? Where? I want to see empirical proof that this is used statistically anywhere in the U.S. before I believe it. Realistically, it can be a non-statistical consideration especially in communities small enough that the judge knows that the last three generations of Smiths out of Dry Gulch Hollow have been drug dealers. Familial relations is also part of the social history given to the judge along with the statistical guidelines. It may be considered, but not statistically.


In the end, this is a philosophical narrative mixed with some fear mongering. The author doesn't seem to have much actual knowledge about the current use of statistics in sentencing and in the end she offers two better uses for all these statistics she seems to think the criminal justice system has easily at hand for every day use. First, rehabilitate the prisoners who are likely to re-offend by teaching them a trade. I'm all for this too. I don't have high confidence in its ability to work in the real world, but it's a wonderful ideal to espouse. Second have friendly officers walking down the street handing out flowers to citizens rather than engaging in proactive policing. Here, I'm also not opposed - partially. Community policing is a good idea although it's usually not engaged in as much as I'd like because police departments don't have infinite resources (and other social and safety issues). However, I'm also a fan of proactive policing. Yes, I know that all sorts of professors have told us all that correlation after correlation after correlation after correlation after correlation after correlation of proactive policing (the dreaded broken windows) with dropped crime rates does not in any way, at all, ever point to any possibility of causation; instead, it's all about aging male populations (because we never make new young males to replace the ones that get old). However, there still seems to be an awful lot of correlation out there.

Anyway, I'll step off that soapbox for now. The use of statistics is generally a good thing. It provides a rational basis for large organizations such as legislatures and administrative agencies to act. However, they aren't being used in the 1984ish manner the author would have us believe. I won't claim this is because governmental agencies are above such a thing. Rather, it is because such a thing is beyond their grasps. I think if the author made a true study of the matter she would be shocked at what she would have to see as rather primitive statistical guideline systems actually in place in the various States and the federal government.