If you get into trouble with the law, Alicia Carriquiry has some advice:
"Before you call a lawyer, call a statistician."
Carriquiry, an Iowa State University statistician and Distinguished Professor in the College of Liberal Arts and Sciences, said when a jury does not understand how statistics play a role in a trial, their improper conclusions can lead to false convictions.
Helping a jury to differentiate a data-informed conclusion from a general assumption is one of the biggest challenges in the justice system. Historically, trained examiners use their own judgment to analyze crime scene evidence. But in many cases, it is shown that forensic examiners overstate their findings, even if their findings are not statistically plausible.
In many cases, it is shown that forensic examiners overstate their findings, even if their findings are not statistically plausible.
When an examiner is linking a bullet found at a crime scene to a suspect's gun, they "can't possibly be right when they say there is 'no way' any other gun except the suspect's could match the bullet found at the crime scene," Carriquiry said. "But the general public does not know that."
To address the problem, the National Institute of Standards and Technology (NIST) has awarded a five-year, up to $20 million grant to establish the Center for Statistics and Applications in Forensic Evidence (CSAFE). Carriquiry is the director of the new center, which is headquartered at Iowa State and calls on leaders in the statistical community from three additional universities: Carnegie Mellon University, the University of Virginia and the University of California Irvine.
In collaboration with scientists at NIST and law enforcement agencies, CSAFE researchers are developing statistical and measurement approaches that will result in scientifically defensible models to quantify uncertainty in certain types of evidence. The center will produce software, reference materials, and apps to use in the field and will provide training activities for forensic scientists.
Quantifying uncertainty
Faulty evidence leads to serious and costly burdens on society. "The Innocence Project," a national organization that tries to exonerate wrongfully convicted individuals through DNA testing and system reform, has helped acquit 337 people since it was founded in 1992. These false convictions are often the result of a lack of a scientific foundation, Carriquiry said. DNA is available in less than 10 percent of criminal cases, meaning more than 90 percent of cases rely on "pattern evidence" (fingerprints, tire tracks, shoe prints, impressions from guns, etc.) and digital evidence (computer files, phone records). Unlike forensic DNA analysis, which has a well-defined measurement process and a biological foundation, most other forms of forensic analysis lack the same degree of statistical rigor.
"If you're on a jury, or you're a forensic analyst, and you know the shoe print from a crime scene is a Nike sneaker, size 11, and the suspect is wearing a Nike sneaker, size 11, it's tempting to say this is a match," Carriquiry said. "But you're not taking into account that there could be several other people in the same neighborhood that have the same shoe. You need to decide whether this piece of evidence is pointing towards guilty or not guilty, and this is where statistics comes in."
Calculating the probability of how many other size 11 Nike sneakers might match this particular shoe print — or perhaps how many cars have the same Firestone tires as a tire track found at a crime scene — is something rarely presented in court.
Calculating the probability of how many other size 11 Nike sneakers might match this particular shoe print — or perhaps how many cars have the same Firestone tires as a tire track found at a crime scene — is something rarely presented in court.
In addition, "It's miserably hard to come up with a robust quantitation of an image," Carriquiry said. "The first step is extracting discriminatory features from an image (such as a shoe print or a fingerprint) to try and find characteristics you can measure. The second step is finding a statistical model for these characteristics. The third - and hardest - step is to try and estimate the frequency of this characteristic in the population."
The center plans to propose an approach that would quantify the match between two bullets, and also develop a method to determine the quality of latent fingerprints (partial fingerprints) in terms of usability in forensic examinations.
Understanding the data
Attaching numerical probability to evidence is one of two primary goals of CSAFE. The other is outreach and training. Examiners need to know how to interpret the data, and lawyers, forensic scientists, and law enforcement need to know how to present the data to the general public. If law enforcement cannot understand the evidence or what conclusions the evidence supports, the center’s efforts will be in vain.
"We always have ideas about how something should work until we read the manual," Carriquiry said. "We need to make sure examiners know how to interpret the data or it won't matter how much work we do."
The College of Liberal Arts and Sciences’ strength in statistics and its tradition of collaboration across disciplines is what will make this project a success. By establishing ties between academic and forensic communities, the center will help create an organized and well-supported research enterprise.
"CSAFE suits Iowa State's land grant mission perfectly," Carriquiry said. "ISU puts a lot of value on research's impact on society, and this will have a tremendous societal impact."
Which faculty inspired you at ISU? Contact Kim McDonough, director of alumni relations for LAS: kmm@iastate.edu or 515-294-7487.