Numbers Meets CSI: Qualifying Value of Forensic Evidence

Forensic statisticians to interpret pattern, impression evidence

Article ID: 625209

Released: 24-Oct-2014 2:00 PM EDT

Source Newsroom: South Dakota State University

  • Credit: Photo by Emily Weber

    Math and statistics department assistant professors Chris Saunders, left, and Cedric Neumann eye a small portion of the math involved in creating formulas that can be used to evaluate forensics evidence. The men received a $780,300 grant from the National Institute of Justice to advance their work.

  • This drawing illustrates how fingerprint features are extracted and organized for statistical analysis. The black dots represent the features, the arrows represent their direction. The gray lines represent the friction ridges on the skin of the finger. The gray letters represent the name of the variables used to represent the features.

  • Assistant professor Chris Saunders of the South Dakota State University Mathematics and Statistics Department

  • Assistant professor Cedric Neumann of the South Dakota State University Mathematics and Statistics Department

Newswise — Nick Stokes of CSI using fingerprints to identify the murdered and Charlie Eppes of Numbers solving crimes via mathematical equations lead many people to assume that forensic science is a highly technical field relying on experts that always have a definitive answer about culpability.

In fact, forensic statistics is a relatively new field that is working to establish investigative techniques and quantitative methods that ensure accuracy in suspect identification. About 25 statisticians worldwide work in forensics science. Two of these happen to be in the mathematics and statistics department at South Dakota State University and have recently received a $780,300 grant to advance the science.

Assistant professors Chris Saunders and Cedric Neumann collaborated to prepare a grant proposal to work on the interpretation of pattern and impression evidence.

At the end of September, they received notice from the National Institute of Justice that their three-year proposal was being funded. It’s the first grant to SDSU faculty from the National Institute of Justice, the research, development and evaluation agency of the U.S. Department of Justice.

Both Neumann and Saunders are relative newcomers to SDSU.

Neumann, a Swiss native, is in his second year in the mathematics and statistics department at State, coming from Penn State after working for the United Kingdom’s Forensic Science Service.

Saunders, a California native, is in his third year in the mathematics and statistics department at State, coming from George Mason University in Fairfax, Virginia, and worked as an intelligence community fellow with the FBI.

It wasn’t by chance that both ended up on the Dakota prairie. The men have known each other since 2006 and formed a bond based on mutual respect for their respective work in forensic statistics. That was what, in large part, led Neumann to leave Penn State to forge new collaborations on the prairie.

Statistical accuracy scrutinized in court
In a February 2012 article in Significance, the magazine of the Royal Statistical Society and the American Statistical Society, Neumann said some courts are scrutinizing fingerprints because of shortcomings in the way the probable value of the evidence is weighed and reported.

Fingerprints have been used for more than a century as a way of identifying criminals. However, fingerprint evidence is not currently permitted to be reported in court unless examiners claim with absolute certainly that a mark has been left by a particular suspect.

“This courtroom certainly is based purely on the opinion of experts, formed through years of training and experience, but not on scientific data. Less-than-certain fingerprint evidence is not reported at all, without regard for the potential weight and relevance of the evidence in a case,” Neumann wrote in Significance.

By establishing the accuracy of likelihood ratios, a statistic used to quantify the probable value of forensic evidence, “courts can begin trusting statistical models when used to report evidence,” Neumann said.

Saunders, whose background has been in supporting FBI investigations, added, “A uniform way of evaluating evidence will make it easier for an agency to decide whether to pursue a suspect or not.”

As Saunders was completing his doctorate in statistics at the University of Kentucky in 2006, he was recruited to do work for the FBI in pattern recognition and handwriting identification. He spent the next two years as an intelligence community postdoctoral research fellow at George Mason.

After the fellowship ended, Saunders continued as an assistant research professor in the document forensics lab at George Mason until coming to State.

In summer 2013, Saunders was a visiting scientist with the FBI lab doing forensics research. “The FBI is trying to build up a group of statisticians because forensics researchers at the federal level understand the need for statistical methods to quantify evidence,” he said.

Research will help law enforcement
The National Institute of Justice-sponsored project, called “ambitious” by one grant reviewer, makes classic DNA forensic profiling seem simple. In fact, from a forensics viewpoint, it is, according to Saunders and Neumann.

“With DNA, there is a well developed probability structure,” Neumann said, explaining that within a DNA sample, the sequencing falls in a handful of ordered patterns. Neumann and Sanders will focus on evidence much more complicated than DNA.

“We will be working on highly dimensional characterization of evidence, where building statistical models is extremely complicated,” Neumann said. “Our objective is to build a solid foundation on which other people can build.”

Saunders said, “This work is more theoretical than we’ve done in the past. This grant funding is to characterize the foundation for that.”

He said they will start with objects that are simple in nature, such as glass fragments from a broken window, where they would look at low dimension characteristics like chemical composition and refraction index. Neumann called that a “baby problem.”

As their statistical models prove accurate, they would apply them to more complex evidence, such as fingerprints, firearms and more complicated chemical data, such as the composition of fibers, he said.

The bottom line is that eventually investigators could use their work to determine the probability that a particular trace recovered at a crime scene (such as a fingerprint or a bullet) was left by a suspect or using a particular weapon.

The outcome being that forensic investigators could use established probability models to evaluate evidence. “If we can establish how good of an estimate the probability value is, it will help the agency or the court trust the forensic evidence,” Neumann said.

“Statisticians now can come up with radically different probability values for the same fingerprint,” Saunders said. “We have to find a way to take raw data and turn it into probabilities. Then forensic experts wouldn’t have to say ‘I think’ and make subjective determinations.”

That could have a “lasting impact on the nation’s criminal justice system,” SDSU math department head Kurt Cogswell said.

About South Dakota State University
Founded in 1881, South Dakota State University is the state’s Morrill Act land-grant institution as well as its largest, most comprehensive school of higher education. SDSU confers degrees from eight different colleges representing more than 175 majors, minors and specializations. The institution also offers 29 master’s degree programs, 15 Ph.D. and two professional programs.

The work of the university is carried out on a residential campus in Brookings, at sites in Sioux Falls, Pierre and Rapid City, and through Cooperative Extension offices and Agricultural Experiment Station research sites across the state.


Chat now!