I love this type of research, but this particular study requires a two-ton grain of salt.
First, these results are based on only 20 total homicide offenders. So when they claim they can classify 75% of murderers correctly, that really just means 15 out of 20.
Second, the final classifier has access to more than 14 variables, including factors like the number of previous convictions, not just brain structure. [1]
If you're fitting 14 variables to 20 data points, it's very likely you're over-fitting. I personally would wait for a much larger study before drawing conclusions.
[1] "The identified variables were age, PCL:YV Factor 1, PCL:YV Factor 2, years of regular substance use, total number of convictions, ICU, SES, brain volume, left and right lateral orbital frontal cortex, medial orbital frontal cortex, anterior and posterior cingulate, right temporal pole, and right and left parahippocampal cortex." http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4055901/
20 total homicide offenders, 135 non-homicide offenders, and two control groups comprising another 41 subjects between them. I agree it's a highly preliminary result, but successfully IDing 75% of homicide perpetrators and 82% of non-homicide subjects is a good start. Why are you leaving out the rest of the test population?
PS I don't think it's perfect by any means, when you look at the population it's pretty particular to New Mexico. Per my other comment, I think the significance of this is that we should be doing a lot more research in our depressingly large incarcerated population to develop a better understanding what makes some criminals dangerous vs others who are just selfish or opportunistic.
I think the key numbers to compare in a study like this are the number of positive labels (in this case, homicide offenders) vs. the number of variables in the model. With n positive labels and n variables, you can almost always find a set of weights that achieve high accuracy. (That applies even though they're using cross-validation -- I suspect the experimenter tried multiple treatments, so it's very easy to trick oneself into believing you've found "the" model.)
For the control set, the model gets 82% accuracy, vs 87% if they had simply classified everybody as a non-murderer. That suggests mild over-fitting to me.
The other two controls are interesting, especially the one trying to match demographic variables, but the sample sizes (n=20, n=21) are too small IMO to conclude much, and I don't think you can pool the control groups since they're drawn from different populations.
That's not to say their conclusions are wrong. But studies on small sample sizes produce spurious results all the time.
The flip side of that observation is that large-scale machine learning techniques applied to relatively "shallow" data sets might produce some very surprising results.
75% of homicide perpetrators identified correctly (15/20) + 82% of non-homicide perpetrators identified correctly (111/135) is a total classification accuracy of 81% (126/155) for the procedure.
But if we just assume that everyone is in the non-homicide group (which is what "innocent until proven guilty" has us do anyway), then that gives us a total classification accuracy of (135/155) 85% for the same sample set. That is, "innocent until proven guilty" is notably more accurate than the system which is being proposed here, and infinitely less likely to deliver a false positive result.
It can't be infinitely less likely to produce a false result or we'd never have miscarriages of justice. Current research estimates the rate of false positives in capital crimes (almost all of which are homicides) at 4.1%. http://www.pnas.org/content/early/2014/04/23/1306417111.full...
But the point here is that chance is only 50%, so if we have a mechanical system that's within shouting distance of matching our extremely elaborate and very expensive trial system then that's noteworthy. 81% is not far away from 87%; what happens when the machine learning system can repeatably identify homicidal behavior with greater accuracy than our criminal justice system? Overall accuracy is highly significant because it will create a lot of pressure to identify murderers in the interests of public safety. Of course we shouldn't rely on such a system because of the risk of false positives, but you'll see attempts to bring it into evidence in more and more cases. If/when such a system starts to outperform the existing on in terms of having a lower number of false positives, then every innocent defendant is going to demand it and every guilty defendant (in the legal sense) will cite it as a mitigating circumstance.
No, they are comparing criminals incarcerated for murder w. criminals incarcerated for something else.
There might exist an (IMHO unlikely) argument that killing someone causes grey matter differences, but the effects of incarceration and such are controlled for.
That's why naive "error rate" is not useful and not used measurement for any classification that's not like 50/50 but more like a "needle in haystack".
A more useful measure is F-score (https://en.wikipedia.org/wiki/F1_score) according to which calling everyone a non-murderer would give 0%, but the method described in article ("81.29% overall accuracy, 80.00% specificity, and 81.48% sensitivity") would give 81%.
two people have the opportunity to invest in 20 startups. the first person says no to all of them, and the second person invests in all 20 of them. all but one go bankrupt. the first person proudly proclaims "i can avoid bad investments with 95% probability!"... and the second is a millionaire.
just pointing out that the "success" of your model depends heavily on how you choose your evaluation criteria. to me, 87% correctly classified seems irrelevant because you can just change the distribution of murderers and non murders and your model would fail. it seems more interesting to me to focus on, say, the rate the model successfully classifies murders.
The risk structure is different. When you make an investment in a company that turns out not to be successful, you wasted your money. If you persecute someone who would never have murdered anyone, you may have done great and irreversible harm an innocent man.
The rate at which it correctly classifies non-murderers is more important for this measure to be useful.
"Here we use neuroimaging and voxel-based morphometry to examine brain gray matter in incarcerated male adolescents who committed homicide (n = 20) compared with incarcerated offenders who did not commit homicide (n = 135). Two additional control groups were used to understand further the nature of gray matter differences: incarcerated offenders who did not commit homicide matched on important demographic and psychometric variables (n = 20) and healthy participants from the community (n = 21)."
Perhaps someone with expertise can comment on a couple of points:
1) Am I correct that this research is preliminary, given the number of subjects involved? It seems too preliminary to even report on HN (to answer my own question), but maybe I don't understand it well enough.
2) Is there reason to believe that the brain differences caused the homicide, rather than believe that committing homicide and getting incarcerated, two very traumatic events, caused changes to their brains? Or that something very traumatic caused them both the crime and the brain changes?
Given a sufficiently high number of input parameters (huge in this study) and a sufficiently small test group (tiny, in comparison to number of parameters in this study), any classifier black-box approach will find some correlation in any arbitrary condition.
1 - All of science is a progress report, subject to change with more evidence. This is one compelling study that will drive more research.
2 - An excellent point easily missed and hard to account for without true longitudinal studies. Early childhood stress is a known agent of brain damage.
In his book "Free Will" (http://www.samharris.org/free-will), Sam Harris discusses the dilemma of what we would do if we suddenly discovered that horrible acts of violence were actually pre-determined by one's genetics and physical traits, rather than by one's (supposedly) controllable will and thought processes.
How do you punish someone who unwillingly commits atrocities, someone who cannot control it and didn't ask to be born that way? What do you do with that individual? IIRC the example was someone committing a murder and then the jury discovering that the person actually had a brain tumor that was making him/her act differently from normal. Do you hate that person? Who do you blame?
In an ideal futuristic society we would be able to alter our genetics in such a way that we do not present an unnecessary threat to others, if we do indeed discover that we're naturally predisposed towards horrible acts of violence.
>How do you punish someone who unwillingly commits atrocities
I look at justice more as a "confinement of the criminal behavior" than "punishment of the criminal". Viewed that way, I don't see a dilemma in being able to better identify the behavior, especially if the confinement is done after a criminal act.
in my opinion more should should share your idea of justice. It's sad that the thought processes behind high-profile criminal justice are typically fueled by the desire to punish, by whatever society at large.
I would argue that not all those with the required traits for committing crimes do not actually go on to committing those crimes, so it would be wrong to preemptively punish them. Instead we should offer something like therapy to help them better control their instincts.
This has been dealt with in court in real life, for cases where someone has killed someone else while sleepwalking, and not had any recollection of the event, or any intent.
As far as I can see, sometimes the killer is acquitted, sometimes is sentenced, and sometimes the sleepwalking counts as insanity, meaning the killer is committed to a secure hospital.
It isn't quite that black and white. Genetics generally predispose you to things, but they don't make you do things. I could be proven wrong eventually, but I don't believe that it's pre-determined that you will commit X crime.
Part of equality is everyone is held to the same standard. So even if it is biologically harder for you to resist falling into a murderous rage, you are still just as responsible if you do.
in a video that i discovered reading the comments on hn, marshall rosenberg argues that the concept and practice of punishment should be abandoned all together. this is a long (3 hour) video, but its actually an amazing point of view: https://www.youtube.com/watch?v=YwXH4hNfgPg&list=FLtlBbsoKI-...
While 75% of homicide defendants may have this defect, the opposite is almost certainly not true (that 75% of people with this defect will commit homicide). This type of research is fascinating, but since we have no way to treat such a condition, its only purpose would be to attempt to identify potential murderers. And that is a very slippery slope (see Minority Report).
incarcerated homicide offenders had reduced gray matter volumes in the medial and lateral temporal lobes, including the hippocampus and posterior insula
...and indeed I've wondered why we don't test for conditions in the incarcerated population that have known potential for altering behavior, eg Toxoplasma gondii see http://wwwnc.cdc.gov/eid/article/9/11/03-0143_article for example.
Assuming for the sake of argument that we could soon reliably identify symptoms of a homocidal disposition, what does this mean for criminal justice and law enforcement? From a purely utilitarian point of view it makes sense to monitor or even confine such people, but from a libertarian (philosophical) point of view, we could not subject them to penal confinement as people can hardly be held criminally liable for deficiencies in their own brain structure, which might be the result of genetics, malnutrition, or childhood trauma.
Some sort of medical incarceration is the obvious alternative, but this is legally problematic. A Supreme Court case O'Connor v. Donaldson (1975) (among others, but this is the big one) made it very difficult to commit someone for confinement in a mental institution without clear evidence that they present a danger to society. Arguably, the US went putting people in mental institutions too casually and without regards for their rights, to a position of not providing sufficient mental health services for all but the most unhinged patients. While this technological advance might one day provide a reliable gauge of a person's homicidal capability, that's a long way from identifying how or when such behavior might manifest.
At a less extreme case than confining someone, should we allow someone with a measurable homicidal tendency to purchase guns? If we forbid them doing so, how do we avoid a tiered approach to citizenship, in which the rights of some are abrogated pre-emptively through no fault of their own actions, but a 'fault in their stars'?
From a different angle, this is also an interesting result for attorneys involved in death penalty defensive work - if MRI scans, which the courts and public are highly familiar with, can reliably predict homicidal characteristics, then the question of criminal responsibility becomes much more difficult to answer, since people have no control over the structures of their own brains.
Well, and this is just a crazy idea here, you could just decide as a society that citizen gun ownership is kind of pointlessly dangerous, with a close to statistically insignificant rate of self-defense usage, and instead move back to a model of an actual well-regulated militia.
We could, but I'm trying to take an 'all other factors being equal' approach so as to minimize the number of assumptions required, and so considered the second amendment issues in terms of how they're currently interpreted rather than how that interpretation might change.
110 years after the adoption of the second amendment, the US 'militia' was defined as 'all men up to the age of 49'. 'Well-regulated' also means healthy and in good order, in the context of the time. This would suggest that all adult American males (up to 49) should engage in regular training to work as militia units - a militia made up of whatever random yahoos show up with no training is not "well-regulated", using either the modern or antique meanings of the term.
In a more positive direction, the state can provide those at risk with counseling and other mental health services to try to prevent the predicted scenario.
A similar idea was attempted in Chicago with those on the other side of this: those with high risk of being homicide victims.
One of the authors of the paper is Dr. Kent Kiehl, who is famous for using a mobile fMRI to scan thousands of inmates in multiple prisons, including many psychopathic offenders.
His book on his research of psychopathy in prisons, "The Psychopath Whisperer", is excellent (despite a terrible name). Easily one of the best books on the topic. I want a sequel.
In recent years research has found more and more evidence of genetic determinism. i.e. certain types of brains are more prone to violent behavior. Psychopaths also have a certain brain "signature" [1]
I wonder if the legal system recognizes genetic determinism as a mitigating factor in violent crimes. If I were a defense attorney in a no-win scenario, I'd probably try this defense. Ah..will probably spend the rest of the night researching [2]
On segment of the show is about a man whose self control is impacted due to some mental disorder. The question is: Should he be treated more leniently because it's not his fault, but his brain's? The conclusion, that I have to agree with is: Of course not! You ARE your brain.
1. Retribution - nope; predisposition to crime doesn't change the appropriate retribution for the act;
2. Deterrence - nope; if we believe in deterrence then we definitely do not want people predisposed to crime to feel that they might get off lightly because of that; those people would be the prime "focus group" of deterrence that'd be threatened to keep their violent tendencies in check or suffer harsh consequences.
3. Rehabilitation - it depends; are the brain factors making you prone to violent behavior curable ? It might change the nature of appropriate punishment from imprisonment towards forcible curing = brain modification; but it's not really a mitigating factor there, more like an agravating one that requires more/different rehabilitation.
4. Incapacitation - it's an aggravating factor in that case; if you're genetically predisposed to violent crime, then according to incapacitation doctrine you should be isolated from society more/longer than someone who is less likely to reoffend.
5. Reparation - nope; no effect.
6. Denunciation - okay, for this theory it would be a mitigating factor. However, denunciation is generally applied to nonviolent crimes (ethics/morals issues); for murder generally we focus on combination deterrence, rehabilitation and incapacitation.
I hate the way the word "abnormal" is thrown around just like that. I would bet that any group of people with a strong common characteristic could be found to have common "abnormalities" in their brain.
<Morgan Freeman>Does this average, normal brain, actually exist? Do the heads of more than 50% of people on Earth contain close to identical brains? Or are we all abnormal in our own way?</Morgan Freeman>
And their tendentious conclusion about predicting homicide is simply creepy.
First, these results are based on only 20 total homicide offenders. So when they claim they can classify 75% of murderers correctly, that really just means 15 out of 20.
Second, the final classifier has access to more than 14 variables, including factors like the number of previous convictions, not just brain structure. [1]
If you're fitting 14 variables to 20 data points, it's very likely you're over-fitting. I personally would wait for a much larger study before drawing conclusions.
[1] "The identified variables were age, PCL:YV Factor 1, PCL:YV Factor 2, years of regular substance use, total number of convictions, ICU, SES, brain volume, left and right lateral orbital frontal cortex, medial orbital frontal cortex, anterior and posterior cingulate, right temporal pole, and right and left parahippocampal cortex." http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4055901/