Bad Science--Confirmation Bias

Reporters Joseph Neff and Mandy Locke at the Charlotte Observer have this terrific look at local "forensic scientists" who are engaged, in their words of one disgusted local juror--fraud.  The story perfectly illustrates the problem of confirmation bias, which crops up continuously in so-called scientific evidence especially "profiling."  Neff and Locke's article follows:

"In the fall of 2007, Davie County prosecutors studied a suspicious bloodstain in a crime scene photograph, looking for evidence that Kirk Turner intentionally slashed his wife's throat with a pocketknife.

Turner, a Kernersville dentist, said he killed Jennifer Turner in self-defense after she attacked him with a 7-foot spear. But prosecutors noticed a V-shaped stain on Kirk Turner's T-shirt. [Dr.] Turner [was charged] with murder, and then turned to the State Bureau of Investigation to help prove their theory: Turner killed her, wiped the knife on his shirt, then staged the scene by ramming the 18-inch blade through his thigh, twice.

Bloodstain pattern analysts Gerald Thomas (pic left) and his mentor, Duane Deaver (pic, above), embraced the prosecutors' theory. Thomas quietly changed his initial report, which was consistent with self-defense. After a renowned bloodstain pattern expert disagreed with Thomas, Deaver and Thomas conducted unscientific tests to shore up the prosecution.

Thomas stuck steadfastly to the story, even after it became clear that he had filed an erroneous account of the crime scene.  Last year, a jury quickly acquitted Turner. The foreman said jurors were stunned by the SBI's conduct. 'Politically, socially, religiously, I'm conservative; I'm a law-and-order man,' said Landon Potts, an insurance claims adjuster. 'But I don't know what other word to use but a fraud.'

Deaver is a major character in the emerging story of the SBI's troubles. His withholding of evidence contributed to three judges in February declaring Greg Taylor, a Wake County man, innocent of a 1991 murder and freeing him from life in prison. Attorney General Roy Cooper then ordered an audit of the blood analysis unit.

At the bureau's crime labs, where Deaver has been a key agent and trainer, analysts charged with using science to solve crimes have hidden test results or concocted bizarre experiments to shore up a prosecutor's case, a News & Observer investigation reveals.  Deaver and Thomas are among at least 10 SBI agents and analysts who have tailored their investigations to please prosecutors, ignored key evidence or locked on to suspects who turned out to be innocent.

For years, bosses have looked the other way. Sometimes they have promoted problem agents.

That may be changing. After questions raised [were raised] Cooper suspended the work of the bloodstain pattern analysts. He also replaced veteran SBI Director Robin Pendergraft with Greg McLeod and ordered him to examine the training, certification and practices of the agents.

But problems with blood analysis have persisted unchecked for years, despite repeated complaints from defense attorneys and damning rulings by judges.  In October, a federal judge found that Deaver gave false testimony in a 1993 capital case when he testified about invisible bloodstains though he had never performed conclusive tests. At a 2004 hearing in that case, Deaver characterized that sort of testimony as routine.

Deaver built the SBI's blood spatter analysis program. He trained its analysts, including Thomas.

Marilyn Miller, a professor of forensic science at Virginia Commonwealth University, said forensic scientists outside of North Carolina have long been concerned about examinations and testimony offered by Deaver and his protégés. But they feel powerless to do anything.  '[Deaver's] not a member of any associations. We can't censure him or demand he change the way he works,' Miller said. 'He trains all these people, and it frightens me to think of the reach of his work.'

Stuart James, one of the world's foremost experts at analyzing bloodstain patterns at crime scenes, was hired by Turner's lawyers to examine the evidence before the trial. Since Turner's acquittal, James has shown the video of the SBI experiments in workshops and conferences in the United States and Europe. Every colleague deemed the work unscientific, James said.

'They thought [the work of Deaver] was a bunch of malarkey,' [Professor] James said.

Deaver and Thomas declined to comment for this report.

Pendergraft, the former director, declined to answer questions about the Turner case.

'I have not studied this case as you have, so it would be inappropriate for me to try to reply to each and every one of your questions,' Pendergraft said shortly before she was replaced. 'Our goal is to be fact finders and to hand the facts to the prosecution.'

[Blood Evidence, Never Lies]

Sheriff's deputies and emergency medical technicians found a bloody mess when they arrived at the Davie County horse farm owned by Kirk and Jennifer Turner on Sept. 12, 2007. Jennifer Turner lay dead in a pool of blood, her throat cut. Kirk Turner was in shock, bleeding from arm, thigh and face wounds.

The Turners had separated and were going through a protracted and bitter divorce. That night, Kirk Turner went with a friend to pick up some tools from a shed on the farm, about 100 miles west of Raleigh. Turner talked with his wife while his friend loaded a truck. Then Turner gave his wife legal papers that included an affidavit from her previous husband and a motion for a judge to compel the sale of the farm.  Kirk Turner said his wife suddenly attacked him with a decorative weapon leaning against the wall: the huge spear. He said she shoved the blade completely through his left thigh, and then skewered his leg a second time, inches from his groin, barely missing his femoral artery.

The defense said he reached into his right pocket, pulled out his pocketknife and slashed at his wife. He slashed her neck twice, in quick succession, and severed her carotid artery. She bled to death in seconds. Police cordoned off the scene. The next day, they called the SBI for a bloodstain pattern analyst.

Understanding a bloodstain pattern can help investigators understand how a violent event occurred. Flat, round drops suggest dripping blood, while tiny droplets suggest blood sprayed from a nearby beating. Scientific studies support some aspects of bloodstain analysis, but experts have cautioned against placing too much emphasis on it, saying it can be biased by the theories of prosecutors or defense lawyers.

Gerald Thomas of the SBI arrived to inspect the crime scene on Sept. 14. He collected photographs and examined bloody clothing. Thomas interviewed fellow investigators, who told him that Jennifer had stabbed Kirk, who then slashed her.

Thomas wrote his observations in a three-page, single-spaced report and attached several pages of handwritten notes and photographs. His report included nothing to contradict self-defense.  But in December 2007, a Davie County grand jury indicted Kirk Turner for murder. Prosecutors had locked on to a theory after they saw Turner's bloody T-shirt.

'It was everyone in our office, all the senior DAs,' said Gregg Brown (pic, right), an assistant district attorney and the lead prosecutor on the case. 'We looked at the photographs, and it appears, gee, that really looks like a knife; I wonder what the SBI will say.'

What the SBI said would have a profound impact on the case. That story emerges from reports, court transcripts and interviews.

[Thomas Changes the Findings to Fit the Facts, per the Prosecutors]

On Jan. 15, 2008, Brown met with Thomas at the SBI lab. Thomas brought Deaver, his mentor.  Thomas and Deaver agreed with the prosecutor that the blood smear on the front of Kirk Turner's T-shirt was made by the knife that killed Jennifer Turner, Thomas wrote in an e-mail message.  There was a problem: In his initial report from September 2007, Thomas had written that the stain was probably made by a bloody hand.  Thomas took his Sept. 14 report and changed one paragraph. He removed 'a bloody hand' and replaced it with 'a pointed object, consistent with a knife.'

Thomas took no notes of the January meeting, though SBI policy requires it. He made no separate report on these activities, though SBI policy requires a report within 15 days. He made no notation that he had amended or changed or updated his report after reviewing the evidence a second time. The date on the changed report: Sept. 14, 2007, the same as the original.

 In April 2009, Kirk Turner's lawyers received a report from their blood pattern analyst, Stuart James of Fort Lauderdale, Fla. James' books on bloodstain analysis are used by the SBI.

James disagreed with many of the agents' conclusions. The smear on the shirt was not made by a bloody hand or a knife, James said. It was most likely a 'mirror stain' created when the shirt was folded, perhaps as EMS technicians cut the shirt off Turner or when they tossed it on the floor.

In an e-mail message to a fellow SBI agent in May, Thomas said he would stick with his knife hypothesis. He promised to conduct tests 'to shore up this argument.'

Thomas and Deaver conducted tests that bore no resemblance to science, at least as it is described by leading experts in the field. Instead of rigorously testing and having peers review their findings, they conducted two tests designed to produce the result they sought.

Video of their work shows that twice, Thomas donned a clean shirt. Each time, he dipped a knife in blood, careful to get blood only on the edges. He carefully wiped the blade on his shirt in an attempt to leave a stain that resembled the outline of the knife. The second attempt was more pleasing to Deaver, who was filming. His comments, audible on the video, were more film director than scientist.  'Oh, even better, holy cow, that was a good one,' Deaver exclaimed. 'Beautiful. That's a wrap, baby.'

Thomas wrote a report summarizing how these experiments refuted the defense expert's finding of a stain caused by folding the shirt.  In an attempt to find out how the T-shirt was handled after it was removed from Kirk Turner's body, Thomas wrote in his report, he spoke with Capt. Jerry Hartman, the lead investigator for the Davie County Sheriff's Office.

'Hartman stated that he was present when emergency services cut the gray T-shirt from Mr. Turner's body and that the questioned blood stain was observed present in its current condition on the shirt. Hartman said he took the shirt from Emergency Medical Services (EMS) and placed it in a secure area (an adjacent room), laying flat on the floor to dry.'

That never happened. At the trial, Hartman testified he entered the crime scene two hours after EMS took Kirk Turner to the hospital. The initial crime scene photos showed theT-shirt crumpled on the floor, inside out.

Later photos show the shirt laid out neatly on the floor to dry.

When Thomas' report about the shirt was shown to be wrong in front of the jury, Thomas refused to change his conclusion, saying he was sticking with his lab experiment.  James, the bloodstain expert, asked several peers to review his work in the case, including Lisa DiMeo, who works for both defense and prosecution as an independent forensic examiner in San Diego. DiMeo said she was appalled by the tests run by Thomas and Deaver.

'I'm glad there was audio, so you could hear the person doing the video directing the actor what to do,' DiMeo said.

Jurors said the most powerful evidence of Turner's innocence was blood evidence.

Turner kept his knife in his right front pocket, which was smeared with blood. DNA tests revealed the blood was only Kirk Turner's. To the jury, that meant Kirk Turner had already been attacked and had his own blood on his hand before he reached into his pocket for his knife.  None of Thomas' reports mentioned the bloody right pocket.

The Turner case is not an isolated instance. In at least three other cases, the tainted work of Deaver or Thomas has played a part in reversed sentences or outright acquittals.

The agents made the same mistakes years apart. In preparing for the 1993 murder trial of George Goode, Deaver ran a screening test for blood on Goode's boots. The test showed the possible presence of blood. The stain was invisible to the naked eye. There's no record that Deaver performed the subsequent test that would verify the presence of human blood.

Yet at trial, he twice unconditionally described the invisible stain as blood. In 2009, U.S. District Judge Malcolm Howard found that Deaver gave false and misleading testimony. Goode's sentence later was reduced from death to life because of ineffective counsel, which included failure to challenge Deaver.

Cooper's chief advisers were aware of Howard's ruling but did nothing about Deaver until February.

At that time, Deaver's work was part of the reason that three judges declared Greg Taylor innocent and freed him from life in prison. The only forensic evidence at the 1993 trial connecting Taylor to the murder was a stain on Taylor's SUV. Investigators testified at trial that the stain was blood. Deaver had run a preliminary test that indicated the stain could be blood. He told police about the results of that test but did not tell them he ran a subsequent test that found the stain was not blood.

Thomas attempted to give similar testimony in a 2008 murder trial in Greensboro. Donna Smith was charged with the first-degree murder of her live-in boyfriend, Walter Norman Little III.  Donna Smith told investigators she had no idea her boyfriend was beaten to death in the driveway until a neighbor called her. She called 911 and gave repeated interviews to police, even after she learned she was a suspect in the case. Police found no evidence of blood in the house.

A year later, Thomas was called in to analyze the bloodstain patterns produced by the beating. He examined photographs of the crime scene and clothing.

On a white T-shirt taken from Donna Smith, Thomas identified a linear pattern, three inches in length, of 14 tiny circular bloodstains. In his report, Thomas said that the stain could be impact spatter, which is made by blood flying with some force, often from the impact of a weapon. Thomas concluded that Smith was not telling the truth. 'Smith denies being near the victim during or after the time of bloodshed. ... The stain observed on the T-shirt belonging to Donna Smith could have been made from impact spatter. Therefore, the bloodstain analysis disputes Donna Smith's statements.'

Before allowing him to testify [and consistent with the rule in Daubert relating to the admissibility of scientific evidence], Superior Court Judge James Webb asked Thomas - with the jury out of the room - what his testimony would be.  'The wearer of that shirt was present during the time of bloodshed,' Thomas said. 'When bloodshed occurred, whoever was wearing that shirt was present during it and close by.'

Smith's lawyer objected: The stains were too small to be seen with the naked eye. More important, the SBI reported that the preliminary test showed only that the tiny stains might be blood. But according to the report, the laboratory did not run the test that would conclusively establish whether the stain on Smith's shirt was blood.  [Judge] Webb ruled that Thomas could not testify. Smith was acquitted. News researcher Brooke Cain and database editor David Raynor contributed to this report.
_______________________________
Confirmation bias is the tendency  to favor information that confirms our preconceptions or hypotheses regardless of whether the information is true We gather evidence and recall information from memory selectively, and interpret it in a biased way.  Our biases show up in a particular way for emotionally significant issues and for established beliefs.  As an example, we tend to have an established belief that the police protect us and have a greater skill at assessing evidence--when a police officer testifies to something he or she perceived we will tend to credit that testimony with greater weight than the evidence given by a lay witness.

Confirmation bias is enshrined in the law, for instance the Ohio Supreme Court say that "trained" police officers can "guesstimate" speeding and courts can sustain convictions based on this testimony.  Likewise the "expert" rule anticipates that greater weight will be ascribed to the "opinions" of persons with "special skill or knowledge in science."  And yet we know that people have been put to death based on "junk science." (see the post on Cameron Todd Willingham,) at least in Texas.

Likewise, we judge one another based on our biases.  I do it.