Tuesday 12 October 2010

Timid Testers and Magic Lassos

Kaiser Fung, of Junkcharts, has written a really nice book about statistical ideas and applications called "Numbers Rule Your World" which I read about half of early Sunday Morning while my beautiful sweetheart was sleeping in to recover from her trans-Atlantic flight. I really like the book, but came across a short passage where I was distrubed by the seemed support for bad statistics.

The issue in question was the use of lie detectors to acquire confessions. "In the United States it is legal to obtain confessions via the reporting of false evidence, which means the police are free to tell a suspect he or she failed a lie detector test no matter what happened." Then he goes on to give an example of when the method had been effective...In the case of the murder of Angela Correa of Peekskill, New York in 1989. There was a 16 year old classmate suspect who met the psychological profile the police had for the killer. There was a problem, though. Physical evidence was either absent, or contradictory... Hair samples did not match the accused, and there were none of his fingerprints on any of the evidence found near the body. But most damning was the fact that the DNA evidence from the semen samples inside the raped victim clearly excluded the suspect. But after all this comes the line, "The police knew they had the right guy..".
so they talked the 16 year old into taking a lie detector test, the least accurate investigation tool (other than a DA looking for a conviction) in the police arsenal. After "eight hours in a ten foot by ten foot room, facing in turns Detective McIntyre and Investigatior Stephens." In the end, the young man made a confession. Fung concludes the section with the sentence, "Without the polygrpaph exam, there would have been no confession and thus no conviction."

I was so troubled by the story, that I went to look up the case, and it turned out that shortly after Fung stopped following the case, (he comments on the young man's release in September of 2006) in November of 2006 the physical evidence led to another confession:
"A convicted killer was indicted yesterday in the killing of a Peekskill teenager, a crime for which another man spent 16 years in prison after being wrongfully convicted. Steven Cunningham pleaded not guilty yesterday in State Supreme Court to charges of second-degree murder and first-degree rape. Mr. Cunningham, who is already serving 20 years to life in state prison for strangling his girlfriend’s sister in 1994, is accused of murdering Angela Correa, 15, on Nov. 15, 1989. He was linked to the crime after new testing matched his DNA to semen found in the girl’s body, prosecutors said. Westchester District Attorney Janet DiFiore convened a panel last week to review the circumstances that led to the wrongful arrest and conviction of Jeffrey Deskovic, one of Miss Correa’s classmates, who was released in September.

1 comment:

Nate said...

Pat,
I found this story today on CNN:
Murder charges were dropped after forensics analysis of the suspect's computer provided an alibi:

"A forensic analysis of Dorian's home computer and its activity put Dorian at home at the time of the shootings, authorities said. The officer also told them he was watching sports programs and using e-mail."

I am dimly aware that there are some pretty sophisticated techniques put into the forensics analysis of a computer, but they must have been pretty impressed to throw out whatever evidence was sufficient to charge a man with murder in the first place.

It also seems to provide that one could write a program that creates alibis. This will be an issue.

I know that my own reaction is a complaint of compromised rigor, but I would not prefer a legal system that demanded notarized documentation or photographic evidence (both of which could be fabricated) to honor an alibi. I might save this for my discussion of type I and type II errors.