This essay began as a comment on the recent Quillette Article Noise: A Flaw in Human Judgment—A Review. It makes for a fascinating read for those interested in biases.
A good essay covering a complex subject. As an interesting side note, when COMPAS, the Correctional Offender Management Profiling for Alternative Sanctions was introduced, it reduced racial bias in sentencing. The software was designed to predict rates of recidivism. However, further statistical analysis showed that sentences for African Americans relative to whites remained higher for equivalent crimes. Why was this?
Unfortunately, the algorithm remains proprietary, so we have to look to secondary analysis of raw data for a general overview of COMPAS. This excellent analysis demonstrates its flaws. However, it would appear that they are missing a vital piece of the puzzle. A while back I was reading a study on the Norwegian claim of a 20% recidivism rate for ex-offenders. Depending on the time frame (which can vary hugely depending upon the sympathies of the institution framing the measurement system) and detection method (arrest, charge or conviction), the analysis found that Norway’s actual recidivism rate varied between 13% and roughly 50% on the basis of these two criteria.
Hidden in the original evaluation, the Norwegians admitted that their system was far less capable of preventing recidivism in gang members as opposed to non-affiliated criminal offenders. This makes a lot of sense when one considers the phenomenon of Hassidic Jews and their historic cornering of the diamond trade. Nobody else could compete because no internal security was required in gem appraisal rooms. The reason was simple- if a Hassidim were to steal from the room and was caught, they would not only be risking criminal sanction and imprisonment, but also complete ostracism from their communities- to the extent that they would never see their spouses again, their parents or their children again. This completely obviated the need for security staff within gem appraisal rooms- a significant comparative advantage. It was only when Indian communities moved into Antwerp that they were able to capture a large portion of the market, one presumes because they possessed similar social sanctions and dynamics which gave them the competitive edge.
This is informative, when one considers why COMPAS may provide such differing results when it comes to race, despite the fact that it represents an improvement on the previous human bias of judges. In the case of gang members, individuals face social sanction and ostracism if they end their criminal associations and behaviour- an added incentive towards recidivism, especially when one considers the strength and intensity of the friendship bonds we form during the teenage years.
This also tells us why judges seemed to be more prone to bias than the COMPAS algorithm. Humans are great at evaluating risks. It’s what stopped us from getting eaten by large predators in our primordial past. But we are terrible at evaluating risks- we have tendency to systemically overestimate some risks whilst discounting other more substantial risks as trivial. If we were more rational, nobody would have a fear of flying and nobody would ride horses. Unfortunately, in terms of Criminal Justice this very human tendency to overestimate some risks made judges biased compared to algorithms.
So what’s the solution to further removing structural bias in criminal sentencing? First, create a far more comprehensive police model for evaluating levels of gang involvement. Create a clear dividing line between those who have a criminal history which was based upon no gang involvement and those who were gang involved, and make a distinction as to the grade of gang involvement. Second, previous research on rape criminology, shows that Dark Triad personality types are predictable perpetuators of Rape Myths, and also show strong correlations as rape perpetrators.
Similar research into corporate psychopaths shows that although psychopaths can be productive in many areas, from cheffing to surgery, they generally make terrible leaders. One key pathology is in the selection and grooming of teams. A normal professional manager will try to divide teams so that they are evenly matched- so that only cooperation, teamwork and hard work will be decisive in competition, as opposed to talent or ability. A professional manager prone to psychopathy or the Dark Triad will instead generate definitive A and B teams- and will then proceed to encourage the ‘A’ team to systemically bully and harass the ‘B’ team, which is obviously suboptimal for the business or institution, because the ‘B’ team substantially underperforms, representing lost capacity. In many ways, the psychopath or Dark Triad types are adept at constructing their own social camouflage through group grooming as a means of normalising their own aberrant behaviour.
In a gang setting we can infer a number of conclusions. First, gang involvement is likely a lure to Dark triad personality types- and they will be heavily overrepresented in gangs, far more so than in corporate management, where their presence is far lower than most would suspect, but still a significant business problem. Dark Triad types will likely be the key instigators of inter-gang violence. They are highly status conscious and have a very high tolerance for risk. They will want to pay any insult in kind, be willing to resort to violence and will likely co-opt others to serve as proxy delivers of said violence. Although they will likely talk much of loyalty, they will possess none, and will likely eliminate internal competitors within the gang with accusations of disloyalty, or by planting evidence and then engaging law enforcement- using either anonymous tips or informers to convey intelligence of incriminating evidence to the police. This also might account for at least a portion of the accusations of police planting evidence.
Ironically, disclosing comprehensive information of this nature might well serve as a great way of persuading gang members to discard their former loyalties, and as a means of somewhat reducing recidivism rates- either as a mandatory custodial requirement for known gang members, or as alternative to prison for petty offences. At the very least, a better understanding of the warning signs of psychopathy within gang culture might serve as a form of PSYOP, programming behaviour away from inter-gang violence, lest individuals who might otherwise call for violence in the event of offence taken, might fear that it identifies them as a potential psychopath to other gang members.
And, of course, using clinical evaluations in the pre-sentencing phase of trial is likely to be a very strong predictor of recidivism risks, especially in terms of violent recidivism. It is a good idea purely on the basis of protecting the public. Such a systemic approach could be key to producing lower sentences for those who this metric predicts will be of less risk in terms of public safety, whilst ringfencing the more violence prone into stiffer sentences. Overall, this sort of approach provides a more humane and safe system, more in keeping with our notions of fairness and justice. In a roundabout way the search for racial justice, if viewed through an empirical lens, can reduce costs to the taxpayer (clinical evaluations being much cheaper than long sentences), limit harms and help identify candidates for reform-based approaches.
Clinical pre-sentencing and psychopath awareness courses for gang members are also far less likely to face the reversals seen with both a more punitive approach (historical), as well as the mounting call for recalls which progressive prosecutors are unlikely to survive unscathed. Like a pendulum, American Criminal Justice seems to scythe from one extreme to the other, rarely pausing in the empirical centre-ground and unheedful of the harms either extreme routinely inflicts. Like most things, it’s all about context, detail and, above all, balance- which is why we should simultaneously restore judicial discretion to judges and arm them with the best empirical decision-making tools possible.
Hello Geary, i wanted to comment but i am late again due to holidays :)
On this subject i can recommend a book called "weapons of Math destruction". Being a technical person myself, i am used to algorithms, programming and so on, BUT, i think this current situation i made to get the life or normal people worse. Are they needed all these "automatic systems to take decisions or optimize resources"? it seems that we can no longer use common sense but do as being told (by the algorithm) , i think this is a bad side effect of a culture where everything is based on blaming-suing-taking to court. The result is that people do not want to be responsible and then the algorithm just give them the right excuse, because being responsible and taking a decision would expose them to loose everything.
BTW on the comments we made regarding the ukranian crisis i wanted to share with you a link about a document of Rand corporation issued in 2019 about weakening Russia, but i lost the link, if you are interested i will fin it.
Apr 7, 2022·edited Apr 7, 2022Liked by Geary Johansen
Humans are fallible and they are influenced easily by factors like full belly, smell and look but also attitude of the defendant and how the facts of the matter are presented.
I suppose you are right that the automated systems may help judges to eliminate some of this bias by proposing the sentences but mandatory or mechanical sentencing is still wrong - the reason we have judges is not only tradition of having the wise and educated in the law (one can hope) making decisions in court but also the simple fact that no set of rules cover it all and in such cases you may want to overwrite the automatic system that just failed to account for all circumstances.
One may argue of course that eventually we will update all the automatic/mechanic systems of judgment with all fringe cases but even if that were possible (which I think it is not) - an attempt to do so will produce a system that nobody understands because it is impossible to understand. We have that already - modern non-algorythmic systems based on e.g. neural networks cannot be understood. The problem of how to see that they did not fail (or how to notice that they did) is not easy to solve.
Besides even judge Dredd was not stranger to changing view on sentencing being influenced by e.g. others less lethal and more compassionate members of the community of judges.
Very interesting. I like your consistently wonky and out-of-the-box perspective.
“ This is informative, when one considers why COMPAS may provide such differing results when it comes to race, despite the fact that it represents an improvement on the previous human bias of judges. In the case of gang members, individuals face social sanction and ostracism if they end their criminal associations and behaviour- an added incentive towards recidivism, especially when one considers the strength and intensity of the friendship bonds we form during the teenage years.”
Hadn’t thought of that but it makes complete sense.
My only pushback is that I’m not sure it would be so helpful to teach gang members specifically about psychopathy and the dark triad as it would to teach them about how some of their fellow members may have betrayed them. If you show people direct evidence that they were betrayed, they’ll be much less likely to remain loyal to the gang. In the abstract, “psychopathic gang members betray their comrades” doesn’t necessarily convince a member that HE was betrayed.
Also, haha, you’re a neocon. You just don’t know it yet. I mean neoconservative, not “neocon” which is used primarily by people who don’t know what neoconservatism was and think that calling someone a “neocon” means calling them a “globalist shill.” (Hint: neoconservatism originally had nothing to do with foreign policy.)
“where their presence is far lower than most would suspect, but still a significant business problem.” - a small number can cause a lot of damage
When learning and discussing psychopathy in abnormal psych -we were taught that IQ plays a role in criminal vs. non-criminal. That psychopaths, like all other humans vary in IQ -the lower are more impulsive, violent, and less capable of long term planning. They tend to get caught more often because of it. Higher IQ are capable to patience ,long term planning, directed goals and weighing pros vs cons and considering outcomes, possible consequences.
So perhaps there are high IQ criminal minds-gang leaders -who use and take advantage of lower IQ -and manage them accordingly? If a new member is a match in intelligence -he will be considered a threat to leadership and be watched closely, possible groomed to rise up in rank or eliminated by a leader threatened.
Very thought provoking. Definitely deeper than what most discuss on the surface.
I’m so grateful that even though QC wasn’t the place for me I am still able to read your writings.
How to Reduce Racial Bias in Sentencing AND Reduce Gang Violence
Hello Geary, i wanted to comment but i am late again due to holidays :)
On this subject i can recommend a book called "weapons of Math destruction". Being a technical person myself, i am used to algorithms, programming and so on, BUT, i think this current situation i made to get the life or normal people worse. Are they needed all these "automatic systems to take decisions or optimize resources"? it seems that we can no longer use common sense but do as being told (by the algorithm) , i think this is a bad side effect of a culture where everything is based on blaming-suing-taking to court. The result is that people do not want to be responsible and then the algorithm just give them the right excuse, because being responsible and taking a decision would expose them to loose everything.
BTW on the comments we made regarding the ukranian crisis i wanted to share with you a link about a document of Rand corporation issued in 2019 about weakening Russia, but i lost the link, if you are interested i will fin it.
Humans are fallible and they are influenced easily by factors like full belly, smell and look but also attitude of the defendant and how the facts of the matter are presented.
I suppose you are right that the automated systems may help judges to eliminate some of this bias by proposing the sentences but mandatory or mechanical sentencing is still wrong - the reason we have judges is not only tradition of having the wise and educated in the law (one can hope) making decisions in court but also the simple fact that no set of rules cover it all and in such cases you may want to overwrite the automatic system that just failed to account for all circumstances.
One may argue of course that eventually we will update all the automatic/mechanic systems of judgment with all fringe cases but even if that were possible (which I think it is not) - an attempt to do so will produce a system that nobody understands because it is impossible to understand. We have that already - modern non-algorythmic systems based on e.g. neural networks cannot be understood. The problem of how to see that they did not fail (or how to notice that they did) is not easy to solve.
Besides even judge Dredd was not stranger to changing view on sentencing being influenced by e.g. others less lethal and more compassionate members of the community of judges.
I seem to remember back in the 60s that they tried teaching criminals about psychopathy, and how to overcome it.
The problem was that they used the knowledge to hide their flaws and become better criminals.
Very interesting. I like your consistently wonky and out-of-the-box perspective.
“ This is informative, when one considers why COMPAS may provide such differing results when it comes to race, despite the fact that it represents an improvement on the previous human bias of judges. In the case of gang members, individuals face social sanction and ostracism if they end their criminal associations and behaviour- an added incentive towards recidivism, especially when one considers the strength and intensity of the friendship bonds we form during the teenage years.”
Hadn’t thought of that but it makes complete sense.
My only pushback is that I’m not sure it would be so helpful to teach gang members specifically about psychopathy and the dark triad as it would to teach them about how some of their fellow members may have betrayed them. If you show people direct evidence that they were betrayed, they’ll be much less likely to remain loyal to the gang. In the abstract, “psychopathic gang members betray their comrades” doesn’t necessarily convince a member that HE was betrayed.
Also, haha, you’re a neocon. You just don’t know it yet. I mean neoconservative, not “neocon” which is used primarily by people who don’t know what neoconservatism was and think that calling someone a “neocon” means calling them a “globalist shill.” (Hint: neoconservatism originally had nothing to do with foreign policy.)
“where their presence is far lower than most would suspect, but still a significant business problem.” - a small number can cause a lot of damage
When learning and discussing psychopathy in abnormal psych -we were taught that IQ plays a role in criminal vs. non-criminal. That psychopaths, like all other humans vary in IQ -the lower are more impulsive, violent, and less capable of long term planning. They tend to get caught more often because of it. Higher IQ are capable to patience ,long term planning, directed goals and weighing pros vs cons and considering outcomes, possible consequences.
So perhaps there are high IQ criminal minds-gang leaders -who use and take advantage of lower IQ -and manage them accordingly? If a new member is a match in intelligence -he will be considered a threat to leadership and be watched closely, possible groomed to rise up in rank or eliminated by a leader threatened.
Very thought provoking. Definitely deeper than what most discuss on the surface.
I’m so grateful that even though QC wasn’t the place for me I am still able to read your writings.
I am a member of the Braver Angels debate committee and plan to quote you in our next debate.
https://braverangels.org/event/national-debate-criminal-justice/