26 March 2013

Objective Impulsivity Test Predicts Rearrest Rate

[R]esearchers studied a group of 96 male prisoners in New Mexico just before their release. Using functional magnetic resonance imaging (fMRI), the team scanned the prisoners’ brains while they completed a computer task measuring impulsivity: They had to quickly and accurately respond when the letter X flashed onscreen, but if it was a letter K, they had to restrain themselves from responding.  
Researchers were particularly interested in a region of the brain called the anterior cingulate cortex (ACC), which is known to be linked to behavioral regulation and self-control.  
Four years after these prisoners’ release, the researchers checked their arrest records. Over half had been rearrested—and their earlier brain scans were telling. Men who had shown lower ACC activity were significantly more likely to be rearrested, and they were rearrested sooner. A man in the lower half in terms of ACC activity had a 60 percent likelihood of rearrest, while those with above-average ACC activity had only a 46 percent likelihood. And the men with low ACC activity were rearrested seven months earlier on average. These numbers held true even after the researchers accounted for other risk factors such as age, drug and alcohol abuse, and psychopathic traits, they report in the Proceedings of the National Academy of Sciences.
From here.

The study used fMRI as its objective test of impulsivity, but this aspect of the study isn't really all that critical, just flashy. There are many less expensive objective tests of impulsivity out there that would serve essentially the same purpose or might even been more definitive, as fMRI intepretation is as much as art as a science and have had less time to have the kinks worked out of them. The "authors themselves stress that much more work is needed to prove that the technique is reliable and consistent, and that it is likely to flag only the truly high-risk felons and leave the low-risk ones alone. “This isn't ready for prime time,” says Kiehl."

The brief news report also doesn't report how material factors like age, drug and alcohol abuse, and psychopathic traits were as risk factors. Ultimately, what really matters is how predictive the entire risk prediction instrument using all of the factors it studies is relative to not using it or making purely subjective determinations. If nothing else in the instrument is measuring impulsivity, yet impulsivity accounts for a quarter of the entire risk and an even greater proportion of the predictable risk, then an objective measure of this trait belongs in the overall risk prediction instrument.

Of course, saying that objectively determined impulsivity really accounts for a quarter of the risk of rearrest, when the measured difference involved twenty-two or so rearrests in the lower half of impulsivity and twenty-nine or so rearrests in the higher half of impulsivity, overstates the findings.  The findings may have been large enough to be statistically significant, particularly given the confirming data on rearrest timing.  But, the study is too small to quantify the magnitude of the effect observed with any great precision.  The odds are good that a meaningful share of the difference could be attributed to random chance.

The end product one is looking for is a set of psychological tests and a set rule for combining their results that can be used when judges and prosecutors make decisions on eligiblity for or recommendations relating to probation, deferred judgment or community corrections program eligiblity, for judges determining sentence length in data included in a presentencing report, and when probation and parole supervision programs and state correction departments make decisions on how intensely a convicted criminal should be supervised in and outside of prison. Individuals at high risks of reoffending should be more intensely supervised and/or incarcerated longer, while those at low risk of reoffending can be lightly supervised, thereby maximizing protections for the public and other inmates with their department's limited budget.

Instruments like these can also be used to demonstrate in an empirically based way as part of a larger presentation to demonstrate how changes in funding for probation and parole programs are likely to impact reoffense outcomes. An ability to identify higher risk individuals with some degree of accuracy demonstrates that available monies are being well spent and that the programs would allocate them to needs that would have a real impact if they had the funds to do so.

Simple point systems that add or subtract points based on predetermined factors in amounts devised to fit a recidivism model of the kind that an impulsivity test like the one used would be incorporated into, consistently outperform in accuracy expert assessments of likelihood of recidivism performed in a more individualized and less structured manner.

Mostly, this is because experts routinely consider facts that empirical studies have determined are not very important, and because expert opinions are frequently fuzzy on the relative importance of the factors that they do consider. Simple point systems devised based upon regression models derived from empirical data address the mistakes of considering too much irrelevant information and giving the wrong weight to the facts that are important better than highly trained experts who use more subjective methods.

While the findings of these studies are commonly described as ways to identify high risk offenders, the real benefit to budgets and otherwise, is to give senior corrections and court system administrators empirical cover for recommendations that individuals whom the point systems show to be "mostly harmless" be lightly supervised. In the absence of such cover, these officials might need to greatly increase the average level of supervision at great cost, while still inadequately supervising high risk individuals and engaging in overkill for a large share of the supervised population.

The study is Aharoni, Eyal; Kiehl, Kent, et al. Proc. Natl Acad. Sci. (2013). Some of the lead author's prior publications in the field include:
* Aharoni, E., Antonenko, O., & Kiehl, K. A., "Disparities in the moral intuitions of criminal offenders: The role of psychopathy." Journal of Research in Personality, 45, 322–327 (2011);
* Aharoni, E., Funk, C., Sinnott-Armstrong, W., & Gazzaniga, M.m "Can neurological evidence help courts assess criminal responsibility? Lessons from law and neuroscience.", Annals of the New York Academy of Sciences, 1124, 145–160 (2008).
* Aharoni, E., Sinnott-Armstrong, W., & Kiehl, K. A., "Can psychopathic offenders discern moral wrongs? A new look at the moral/conventional distinction.", Journal of Abnormal Psychology (2012).

Another of his papers published online before print earlier this year correlated scores on a commonly used psychopathy evaluation instrument, the Hare Psychopathy Checklist–Revised, against confidential reports by offenders in prison of prior unsolved violent crimes committed, and determines that psychopaths are far more likely to have committed previously violent crimes for which they were not caught and that some parts of the checklist were correlated with more undetected prior crimes (antisocial lifestyle and behavioral psychopathic traits) while others (affective psychopathic traits) were associated with fewer undetected prior crimes relative to others with the same psycopathy score.

Indeed, this result, taken together with the rearrest data, suggests that part of the reason that the rearrest rate shows a fairly strong relationship to impulsivity, is that impulsive people not only reoffend, but get caught doing so, while less impulsive but pyschopathic parolees might be reoffending just as often but not getting caught as often.

The frequency of psychopathic traits in the general population is similar to the incidence of bipolar disorder or schizophrenia, for example, but these traits seem to be somewhat more of a matter of degree than those forms of psychosis.

No comments: