Fresh back from trademarking the Sentence-O-Matic 1000, Jake DiMare proudly advised me that this idea wasn’t dead yet, as demonstrated by Judge Richard Kopf’s continued push toward sentencing empiricism. Oh, that Jake. Oh, that judge.
While yanking Attorney General Eric Holder’s chain for wanting his cake and eating it too, by decrying the use of race as a predictor of recidivism while prisons are filled to the brim with men of color put there on Holder’s watch for living the lifestyle of poor black rather than Attorney General (or president) elite black, the judge proffers an alternative list of factors as predictors. Notably, race is crossed off the list.
Let’s face it friends, “race” is low-hanging fruit. It is too easy to attack, although the social science data on race when used as a predictive metric for sentencing is not really about genetics (“race”) as a causative factor in crime. It is about being correlated with crime, and there is a huge difference between the two (causation and correlation). But the word “race” is too freighted with the notion of “discrimination,” so let’s just agree for the sake of argument that empirical data on “race” will never be used at sentencing.
And indeed, race is certainly a flash point, despite its utility in what passes as empiricism these days. So here’s the balance of the list, taken from James C. Oleson’s article on data-driven sentencing.
1. Criminal Companions
2. Criminogenic Needs
3. Antisocial Personality
4. Adult Criminal History
5. Race
6. Pre-Adult Antisocial Behavior
7. Family Rearing Practices
8. Social Achievement
9. Interpersonal Conflict
10. Current Age
11. Substance Abuse
12. Family Structure
13. Intellectual Functioning
14. Family Criminality
15. Gender
16. Socio-Economic Status of Origin
17. Personal Distresss
These 16 factors create a construct to empirically predict who will go out and commit crimes. It takes little effort to deconstruct these factors in terms of race, poverty, educational opportunity, future prospects. These are all proxies (except gender) for race, even though race is crossed off.
To the extent that Judge Kopf calls Holder intellectually dishonest, there is no doubt that’s true. Judge Kopf concludes that Holder is an “ostrich,” hiding his head in the sand when calling for the elimination of race as a consideration, while using every manifestation of racial prejudice available to justify the same outcome.
I suspect that Holder is hardly an “ostrich,” but only because I don’t believe the attorney general isn’t well aware of what Judge Kopf is saying. The AG knows that the system is replete with proxies for racial disparities that impact blacks disproportionately and perpetuate those disparities. If daddy is in prison, son will be there soon enough. But if we don’t mention that daddy is black, then it’s racially neutral, right? Run the numbers, guys.
As I responded to Jake’s irrational exuberance over the billions he plans to make with the Sentence-O-Matic 1000, this is still just pigeonholing people. As was noted in the comments to the judge’s post, these are all factors that are subsumed in sentencing now under §3553(a), to the extent they apply. There is no limit to what is thrown against the wall to see if it sticks.
And it’s completely understandable from the perspective of a judge, the person whose job it is to fashion the “right” sentence, that there be concrete parameters so that it’s easier to stick each defendant into a pigeonhole. It makes for a far cleaner mess. Who doesn’t want to believe they’ve done the right thing because 16 factors say so? No more nagging, sleepless nights pondering lives ruined. The Sentence-O-Matic 1000 says you gave the right sentence, and so you did, dammit.
As an aside, the notion that data-driven sentences will somehow result in a reduced prison population eludes me. Using these factors to determine recidivism is neutral. They’re just factors. They no more suggest a greater or lesser number of people going to prison; they merely provide a comparative basis between defendants.
But will the comfort of knowing that there is some data-driven basis to justify a sentence change how long or short an individual sentence will be? Not unless somebody decides that life plus cancer isn’t the baseline sentence for anyone who commits a federal offense. But that’s a different problem, and I digress.
Where in that list of predictive factors is there anything about a teenager who has to fight his way down the street to survive, because he lives in a tough neighborhood where nice boys end up dead? Where in that list is there a chance for the child of a drug-addled parent to go to college and become a nuclear physicist? Where in that list does a kid with a pot bust go on to become president?
The data says these things won’t happen, and they most assuredly won’t as long as we’re playing the odds. These are the same odds that inform a judge that the cop on the stand is more likely right than lying, and the defendant in the dock is more likely guilty than railroaded. Because of these odds, lying cops get a pass, and innocent defendants get convicted. And kids who grew up under unpleasant circumstances not of their making are more likely to be recidivists.
Yes, yes, yes. The odds, the data-driven odds, are all probably true. I concede the point. But they are not absolute. They are not immutable. The rare person can break free of the odds and go on to create great music, beautiful art, a cure for cancer. But when concern over facilitating a pseudo-scientific approach to sentencing becomes more valuable than taking a hard individual look at every human being standing before a judge for sentence, we are guaranteed that they will never climb out of the pigeonhole into which they’ve been dumped.
If we allow ourselves to fall into the facile methodology of locking in our worst factors because the data tells us that these will keep the recidivists off the street, we condemn the best along with the worst. We freeze the status quo, and yawn as generations repeat the failings of their predecessors because the data refuses to give them the chance to be better.
There is something warming, comforting, about removing the burden, the responsibility, for destroying people’s lives by using empiricism. It’s no longer the judge’s fault for depriving a human being of the chance to break free of bad parents, bad neighborhoods, bad schools and, yes, racial prejudice. The data made him do it.
And I still fail to see how this means we end up with fewer people in prison. It just means we feel better about putting them there.
Update: Judge Kopf posts an email from Dr. Oleson, which includes this:
People dislike the “sentence-o-matic 1000” but rejecting algorithms and automation does not prevent assessments of risk – it just means that the human estimates are likely to be more idiosyncratic and less accurate
Dr. Oleson is clearly correct that they are likely to be more idiosyncratic. As to less accurate, that depends on whether the rejection of algorithms and automation means we remain constrained by the pigeonholes or expand the factors to everything, data-be-damned, that reflects the full panoply of human life, character, nature and circumstances, no matter how varied they are in any case.
Mind you, for those lawyers who fall back on the routine sentencing arguments rather than provide the court with insight into your client’s entire life, world and circumstances, this won’t matter. You’ve pigeonholed your client already. For those judges who only care about the shallow, pedestrian factors, same thing.
For those who remember that a human life is at risk, and strive to do better, to provide a full and rich picture of the person before a court, those idiosyncratic estimates are what distinguishes humanity from machines. As for accuracy, who are we kidding?
SHG,
You write: “And I still fail to see how this means we end up with fewer people in prison. It just means we feel better about putting them there.” I must not have been clear as I think there may well be fewer people in prison and that about money, not principle.
AG Holder is on the bandwagon to reduce prison populations particularly federal prison populations. Prisons are like the Defense Department–they eat money the way wolves eat meat. So, there will be the inevitable push for federal judges to find “alternatives to corrections.” That in turn means sorting out who stays in prison and who goes out into the half-way houses and streets with those tre’ chic ankle braclets.
How does the judge sort who goes where? Of course, keep the recidivists and violent in prison, and let the others out in the community. If those are the marching directions given to federal judges, then we can sort in one of two ways.
We can guess. We can use good social science data to more accurately (but never even close to perfectly) predict who will be carjacking someone ten minutes after walking out. Holder is intellectually dishonest because he knows this is the choice. He just doesn’t want to admit that reducing prison populations is not going to be a Kumbaya moment. No fingerprints. Leave it to the judges.
All the best.
RGK
Here’s where the point gets lost: First, nothing compels a judge to sentence a person to prison when prison isn’t required by 3553(a). Nothing is stopping a judge now from “reducing the prison population.” By introducing data-driven sentencing, it’s unclear how that won’t give rise to an even greater risk assessment for more people than a lesser risk assessment. Data takes no sides. Data doesn’t compel lower sentences.
Data is just data. What is done with the data is what matters, but judges already have that ability. They always have, and chose to exercise it by putting ever more defendants in prison. Will data-driven sentencing move the bar from long and shorter sentences, or long and longer sentences?
If reducing the prison population is a goal (which is good with me, as you know I consider the sentences meted out outrageously harsh and sever), then stop sending everybody to prison. Take the 60 month guy and give him probation. Take the 121 month guy and give him 36. Take the 181 month guy and give him 60. You don’t need data to change the trend.
I assume you meant that rare people “can”, rather than “can’t”, break free of the odds to do amazing things.
It seems like this is an intractable problem, how to use data to ensure “appropriate” sentences without removing the human aspects which, ideally, can assess proportionality and can exercise mercy when data suggests a sentence that does not fit either the crime or the offender.
Yes, you’re right. Fixed.
You besmirch me sir. My irrational exuberance is caused by far less odious, yet possibly more embarrassing motives than profit: I simply enjoy the attention to my idea, which I have learned from reading Dr. Oleson’s papers is hardly original, or even new.
However, what I’ve also learned this week is the Sentence-O-Matic 1 already exists. It just exists in ‘MeatSpace’ -A platform for processing information which is rife with underlying subroutines notorious for impacting logic. Like the ego that drives me to continue trying to make my point in this conversation I am clearly, woefully ill-prepared to participate in.
More specifically, the meat-space problems impacting the Sentence-O-Matic 1 (SRA) include:
1. It’s too harsh. For whatever reason the sentencing commission are really really tough on crime.
2. It’s too complicated. In terms of practical application, the interface, a series of giant look up tables and a novel of rules, is causing all sorts of output errors.
3. For a while the sentencing ‘guidelines’ were believed to be mandatory. This problem has been corrected by ‘Booker’ (I think)
Problem #1 is not something Sentence-O-Matic 1000 will correct. You’re going to have to call your representatives in congress, or the sentencing commission, or possibly pray to the deity of your choice about that problem.
Problem #3 seems to be solved at this point.
However, problem #2…Scott…We have a winner. This…This is a problem computers are really, really, awesome at solving.
So, let me re-frame a question I asked before, with my new knowledge:
The scenario is you represent a guilty defendant who is about to be sentenced on statutes which are notorious for coming back with bad sentencing because the combination of factors are well-known to produce confusing, conflicting, or un-fair results using the old-fashioned SRA lookup tables. In this situation, would you prefer:
a. The judge uses the old fashioned look up tables.
b. The judge uses an automated system well known for reducing the time & complexity of the process while always producing the expected results.
Final mitigating factor: The judge will still, of course, have their discretion once they are done determining what the guidelines recommend.
Under these circumstances, which would you prefer?
Look up tables? You’re making my head hurt. Because the guidelines (or as you call them, look up tables) were awful doesn’t mean it’s a virtue to trade one awful methodology for another.
Have you considered just creating a computerish looking box that emits a random number plus a smack, in a stern voice, like “121, you miscreant”? It would have all the same merit with a lot less work.
Empiricism is a fine concept, but if it’s applied to answer the wrong, or maybe the not-quite-right, questions, well, then it’s just empiricism for the sake of empiricism . . .
Where are the randomized controlled trial(s), literally? Meaning, wouldn’t it be interesting if experiments could take a randomly but statistically significant number of “the 60 month guy and give him probation. Take the 121 month guy and give him 36. Take the 181 month guy and give him 60.” while doling out to the control group the current (and harsh) sentences and then follow the outcomes to answer the right questions, like recidivism rates and short and long-term impact on lowering prison populations. For example . . .
Or how about experiment with other variables that might be causal factors to lower recidivism rates and other positive outcomes (as opposed to correlative factors to negative outcomes) — maybe one group gets intense therapeutic intervention while in prison while the control group gets what they currently get, which probably isn’t much; or perhaps the variable of one group being trained and given skills with while in prison and placed in meaningful jobs upon release while the control group is left to fend for themselves, as they currently do . . .
I guess I mean it seems like we have a large and statistically significant enough supply of convicted criminals to perform all sorts of useful experiments. We have nothing to lose, do we?? . . .
Your second paragraph is on target, but it goes further. How about educate them first? How about let their daddies out of prison? How about give them decent jobs to earn a decent legal living so they’re vested in the system?
The only thing prison brings to the mix is prison. How about an empirical analysis of a camp versus secure prison, where they learn how to be better criminals while sitting in SHU, getting more mentally ill and aspiring for hopelessness?
Recidivism isn’t a product of the length of sentence, but the entire experience. Yet, all judges can do is impose a sentence of imprisonment.
Yes, the possibilities are endless, yet the will appears limited. Though I might point out, and as you well know, it is easy to imagine boiling the ocean — while in reality, we are fortunate to boil a cup of water (or soup) . . .
I guess conceptually I was thinking those things that might be isolated and tested at the point where a convicted criminal lands on the judge’s doorstep. . . .
I think your point was more expansive. If we want to prevent recividism, we can (a) give longer prison sentences, or (b) change the things that cause or perpetuate recidivism. The former is easier. The latter more effective and specifically applicable.
Mr. Greenfield,
Is the likelihood/possibility of recidivism a legitimate basis of sentencing in the civilian justice system? I must admit that this goes way beyond my experience.
Pete Brownback
Absolutely. Specific deterrence is a basic sentencing justification.
“for those lawyers who fall back on the routine sentencing arguments rather than provide the court with insight into your client’s entire life, world and circumstances, this won’t matter. You’ve pigeonholed your client already.”
Ouch – you touched a nerve there. Mitigation (as opposed to sentencing guidelines) is something we prefer not to think about (Ok – *I* prefer not to think about it. I still feel the need to hide behind an amorphous we). Mitigation as a subject is not taught much outside of the capital defense realm. I have umpteen books on cross-examination, trying cases to win, suppression (I mean search & seizure), etc. and not a one on mitigation. And I think most of us would rather identify with Rumpole (“Never Plea!”) than with Guthrey Featherstone, QC (“We should work on mitigation”). Mitigation means we’ve already lost.
Yet, there’s no escaping the fact that many more of my clients will be sentenced than will be acquitted after trial. And it’s not about winning and losing, but helping clients at every stage in the system. You have written in the past about falling into the trap of the Gerry Spence never-lost-a-case mindset (“being me means I never have to prepare a mitigation argument”), but your post just reminded me that Clarence Darrow’s most famous case (Leopold and Loeb) was only about mitigation. And his work there was beautiful.
Maybe there’s a reason for non-capital defenders such as myself to go to one of those conferences. At the very least to seek out opportunities to learn about this non-exciting, depressing, and absolutely vital stage of our work. Thanks for the kick in the butt.
There are far too many CLEs on how to lose rather than how to win, for my tastes. That said, there are far too many on how to market rather than how to practice, so there.
But rather than think of it as mitigation, sentencing is just another part of the job, and once a plea happens, you shift focus to mitigation. Pre-Booker, it was all about guidelines, notes and heartland, but now that 3553(a) is alive, it’s all about the “no longer than necessary.” Whether you call that mitigation or just sound sentencing argument isn’t important.
Some years ago, I taught a CLE on pre-sentence memoranda. I am not a fan of sentence mitigation “specialists,” who do cookie-cutter jobs that have no impact. It’s got to be real, personal and sui generis. I’ve met evil people, but no one who was so evil that there was nothing worthwhile to present to a judge. It’s out job to present it, and do so persuasively.
It strikes me as unfair to base the sentence on factors the accused can not influence. He certainly can influence his own criminal history, but not gender nor family rearing practices. Your parents abused you and beat you too much, so we will give you harder punishment? Seems unfair to me.
What about combinations? Aren’t bad rearing practices and antisocial personality proxies for criminal history anyway? It is honest question, I do not know what the stats are.
This isn’t about fair, but about predicting the future. The stats all suggest that the 16 factors are good predictors. The problem is that they condemn a whole lot of people to lengthier prison sentences for things they can’t control, things they have yet to do, and deprive them of the chance to be better than the stats predict.
While empirical evidence may suggest that a “reasonable person” with the 16 various scores will be likely to reoffend, judges don’t sentence a “reasonable person” with only those characteristics, judges sentence an individual. Leave the “reasonable person” over in torts, where it should be
(Which, since that is a standard for certain civil cases, wouldn’t that be a place where the empirical evidence could be used, since a “reasonable person” would act within a certain range of actions under given conditions [that have been empirically studied]?)
The only possible way to justify the use of these factors would be if we switched the entire focus of prison to rehabilitation. If we could offer education, therapy, skills that lead to productive employment post-release (and a general sense of human dignity) in prison we could address many social problems that lead to generations of families cycling through our prison system.
Instead of using these factors to determine the risk of an individual reoffending in order to justify locking him in prison for longer, why not use the factors to determine the best course of action to stop that individual from reoffending?
Why not use these factors to change the individual in a positive way, making him less likely to reoffend, instead of using these factors to nurse judges’ consciences after making the decision to sentence an individual from a poor family with criminal relatives to a longer sentence than he would a well-off kid from a stable home that commits the same offense because the “empirical research” suggests that a “reasonable person” with those characteristics is more likely to offend?
Any formulaic, factor-based research cannot account for the millions of possible mitigating circumstances about the offense and the individual. The quest for a cleared conscience cannot be devoid of the consideration of an individual as just that – an individual.
Judges may forget that they are sentencing human beings; and, maybe that they are human beings, too. A sentence of twenty years rolls easily off the tongue, and afterwards the now-inmate leaves the court and the judge rarely hears from him again. The papers get filed, and the judge moves on with his life, as he must in order to do his job. After so long it is probably exhausting considering so many factors about so many different people and crimes. But when a judge starts seeing a blank-faced, “reasonable person” with pre-determined factors* during sentencing instead of an individual, maybe it is time for that judge to get a different gig.
*many of which are not even within the individual’s control.
Well said.