CSAFE All Hands Meeting Poster Presentation Summary

The posters at the All Hands Meeting were interesting and informative projects. There were also people from various fields in forensic science and statistics. One poster, authored by Karen Pan, a graduate student, and Karen Kafadar, a commonwealth professor and chair, at the University of Virginia, was about “Quality Metrics for Pattern Evidence”.¹ Their main research question was if they “were given a latent fingerprint, can [they] use the fingerprint’s quality to determine the probability that Latent Print Examiners will find the right match?”. They based their observations off of data retrieved from one database, the Peskin and Kafadar database. They are currently working on getting permission from three other databases, DFIQI, LQM, and SNoQE. The further direction or one of their goals is to combine the four databases’ statistical methods to achieve one method of determining the quality scores of the latent fingerprints. This is beneficial because then there will be a standardized method to determine quality scores of latent fingerprints, which yields more clarity.

They do not necessarily have any important results because they have not figured out a way to combine the databases. One aspect of the presentation that was concerning was that the presenter had no experience in fingerprinting at all. She did markups on one fingerprint, but she admitted that she did not know exactly what she was doing. Also, a fingerprint analyst had a lot of questions and concerns for the presenter. The analyst asked questions about how to count minutiae on fingerprints. She also asked about how long Kafadar had been researching the topic. Pan did a good job of answering the questions, but she would have been better prepared if she knew more about fingerprinting. She could have taken a course or done better research on fingerprinting mark-ups. This is the dilemma that was discussed earlier in the day where everyone was talking about how forensics students need to have a good background in statistics as well as statistics students having some knowledge in forensics. This would do a great justice in the way the research is done. It would also give the researchers and presenters an appreciation of forensic science as a whole.

Another poster, authored by Amy Crawford, a graduate student, and Dr. Alicia Carriquiry from Iowa State University, was about handwriting analysis. The title of the poster was “Statistical Analysis of Letter Importance for Document Examination”. ² Their main research question was about seeing if they could take an automated approach to statistically assess whether a piece of questioned writing lies within the natural variability, or natural fluctuations, of any one of their writers. They also wanted to see if the writing was outside of the other writers’ natural variabilities. Data was collected by having nine people write paragraphs and then the Flash ID® software would break the words into smaller pieces to see how the nodes, a point where three or more lines cross, connected. The researchers calculated the relative frequencies of occurrences for each writer. The results showed that there was a .50 probability that authors of the documents matched. The presenter said that the data was not reliable because the study observed only a small number of handwriting samples. She said that they would have to study the handwriting of more people as well as collect handwriting samples from other databases. An important result was that they found out that the origin of a couple of letters were easier to determine than others proving to be informative to document examiners. The letters were “h” “I”, “e”, “o”, and the cursive “i”.