Scientific American reports, "Who Will Die? Computer Predicts Which Death Row Inmates Will Be Executed."
Capital punishment is legal in 36 states, but that does not necessarily mean all of the condemned will be executed. Some will languish behind bars for life and others may actually be exonerated and set free. Now researchers say they have built a computer system that can predict with 92 percent accuracy which death row inmates are most likely to be executed, a development they hope will lead to a fairer appeals process.
According to the system, the death row inmates most likely to be executed are those with the lowest levels of education. The researchers, from Texas A&M University–Texarkana and Loyola University New Orleans, report in the International Journal of Law and Information Technology, that neither the severity of the crime nor race—the latter of which is often cited as a key factor in convictions—are reliable forecasters of a prisoner's fate.
The system consists of 18 computer processors designed to analyze data the way that a human brain does—by studying one set of data and comparing it with another data set to find similarities and differences. In this case, researchers fed the system information about 1,000 death row prisoners, including their sex, age, race, highest year of school completed, the state in which they were incarcerated, and whether they were ultimately executed or spared. Once the system had established patterns (of traits most prevalent among the executed) from this initial pool, the researchers fed it similar information about 300 more prisoners (leaving out whether they had lived or died). The system, using logic it had developed from the first set of data, correctly predicted the outcome for 276 (92 percent) of the prisoners.
The system's success "has serious implications concerning the fairness of the justice system," says Stamos Karamouzis, dean of Regis University's School of Computer and Information Sciences in Denver, who led the 2006–07 study when he was a professor of computer and information sciences at Texas A&M. "People against the death penalty use the results of this work by pointing out that the nature of the crime has nothing to do with whether you're executed or not."
And:
One legal expert says that it is more important to determine how the death penalty is meted out during the sentencing phase of different cases as opposed to predicting who will be executed once sentenced (which is what Karamouzis and Harper analyzed). Once a person is given the death penalty, "who gets executed is fairly easy to predict," says Richard Dieter, executive director of the Death Penalty Information Center, a nonprofit organization that collects and disseminates information about the death penalty to the public.
The best predictor, he says, is the state in which a person is convicted: There were 42 executions in the U.S. last year, 26 of which took place in Texas. "In states like Texas and Virginia, chances are your (death penalty) case will be upheld, and you'll be executed," Dieter says. "In California, chances are you'll get a reversal during the appeals process or you'll be in jail until you die."
The original article, "An Artificial Intelligence System Suggests Arbitrariness of Death Penalty," by Stamos T. Karamouzis and Dee Wood Harper is available in Adobe .pdf format here. It appeared in the International Journal of Law and Information Technology.
The abstract:
The arguments against the death penalty in the United States have centered on due process and fairness. Since the death penalty is so rarely rendered and subsequently applied, it appears on the surface to be arbitrary. Considering the potential utility of determining whether or not a death row inmate is actually executed along with the promising behavior of Artificial Neural Networks (ANNs) as classifiers led us into the development, training, and testing of an ANN as a tool for predicting death penalty outcomes. For our ANN we reconstructed the profiles of 1,366 death row inmates by utilizing variables that are independent of the substantive characteristics of the crime for which they have been convicted. The ANN’s successful performance in predicting executions has serious implications concerning the fairness of the justice system.
Comments