Posted May 5, 2026
Researchers from the Center for Statistics and Applications in Forensic Evidence (CSAFE) will present at three upcoming conferences. Those conferences are the American Causal Inference Conference on May 11-14 in Salt Lake City, Utah, the American Society of Crime Laboratory Directors (ASCLD) Symposium on May 17-21, and the Association of Firearm and Tool Mark Examiners (AFTE) Annual Conference on May 24-29.
CSAFE researchers will share the latest advancements in their research activities using statistical methods. Below is the list of presentations.
The American Causal Inference Conference
Causal Inference and the Limits of Forensic Evidence
Plenary Session: Evidence on Trial: Causality in the Courtroom
May 12, 10:30 a.m.-Noon MDT
Maria Cuellar, assistant professor of criminology at the University of Pennsylvania
Causal inference is central to legal decision-making, yet the data and methods used in forensic and medical evidence often fail to support causal claims. Focusing on Shaken Baby Syndrome/Abusive Head Trauma (SBS/AHT), this talk shows how common approaches rely on data without ground truth, conflate prediction with causation, and condition on variables that may introduce bias. Cuellar will demonstrate how different analytic choices can lead to fundamentally different conclusions from the same data and argue for a more explicit causal framework to improve the scientific validity of evidence used in the law.
ASCLD Symposium
Sampling SWGDRUG Meets ENFSI Meets CSAFE – Getting the Horse Before the Cart
Plenary Session: Back to Basics
May 19, 11:30 a.m.-Noon EDT
Hana Lee, CSAFE postdoctoral research associate, and Jeff Salyards, CSAFE senior scientist
Forensic Science Service Providers face special challenges when they receive a submission that has a large number of similar items, like pills or bindles. Our normal default position of thoroughly testing every item submitted breaks down in the face of receiving 1,000 capsules in a single case. Over the years, SWGDRUG and ENSFI have provided guidance about how to create a sample from the original submission. These approaches use the hypergeometric distribution and allow examiners to make inferences about the larger whole population from the results of a smaller sample. Various manufacturing industries also employ the hypergeometric distribution to create samples for quality control. This presentation will review the strengths and weaknesses of the SWGDRUG and ENSFI approaches. We will also explore how the quality control use case in industry is different from the investigative use case in forensic science. Finally, we will present an open-source CSAFE tool that will help FSSP leadership make decisions about the best policies and procedures for creating samples.
Meeting Minutes: How to Spend Quality Time with Your Team
May 21, 11-11:30 a.m. EDT
Jeff Salyards, CSAFE senior scientist
As we progress into bigger and bigger leadership roles, our time is spent more and more on meeting with people. Many of us worked hard, early in our careers, to develop expertise in our forensic science disciplines. Fewer of us, later in our careers, think about developing excellence around how to run a great meeting. Leadership authors like Patrick Lencioni, Ken Blanchard, Spencer Johnson, and Steve Covey offer some valuable insights and suggestions. The key idea is that there is no such thing as “a” meeting; there are about nine different interactions we have with our teammates. Although each one gets called “a meeting,” each has a different purpose and requires different skills and preparation on the leader’s part. Imagine a workplace where your teammates actually look forward to meeting with you and each other!
AFTE Annual Conference
Close Non-Matches and Database Searches
May 25, 5-7 p.m. EDT and May 26, 4-4:30 p.m. EDT
Blanca Parker, graduate student at Iowa State University, and Alicia Carriquiry, CSAFE Director
Suppose a crime has been committed and firearms examiners have recovered spent bullets, cartridge cases, or fingerprints from the scene. To generate investigative leads, examiners often rely on proprietary systems such as NIBIN and AFIS to correlate a piece of evidence to potentially similar bullets or fingerprints recovered from other sources. Investigators upload images of the crime scene item(s) onto the system, request a list of the top k most similar items from the database, and then decide whether any of the k candidates warrant a physical comparison.
But how reliable are database searches? Suppose that one match is in the database. How likely is the true match to be included in the list of the top k correlations?
To explore this question, we carried out a simulation where we varied the potential number of correlations (N) and the number of candidates (k). We fit beta probability models for same (SS) and different source (DS) bullets based on similarity scores calculated from the Hamby-44-CSAFE dataset. To simulate the search of a fingerprint database, we use logistic distributions found in Swofford et al.
We considered databases of size N = 500, 5,000, 50,000, and 500,000, and for each, requested the top k = 10, 20, 50, 100 most similar items when N-1 items were sampled from the DS distribution and 1 from the SS distribution. For different distributions and types of evidence, we see the same general trend where reliability decreases as database size increases.
From the Blog
Recent Posts
Keynote Speakers Announced for ICFIS 2026
August 10, 2026
The Scientific Committee for the 12th International Conference on Forensic Inference and Statistics is pleased to announce the keynote speakers …
CSAFE Researchers will be Panel Speakers at NAFSB 2026 Conference
August 7, 2026
The annual National Association of Forensic Science Boards (NAFSB) Conference will be Oct. 13-16 at the Milwaukee Crime Lab Bureau …
ICFIS 2026 Moves Online, Abstract Deadline Extended to Aug. 17
July 27, 2026
The 12th International Conference on Forensic Inference and Statistics (ICFIS) will now be an online-only event, allowing participants from around the world …