I'll be attending the American Society of Criminology Conference in LA this week and came across this session:
QuoteEvaluating Voice Stress Analysis Software for Deception Detection
Sponsor:
Category XXI: Measurement and Methodology
Issues in Evaluation Research
Schedule Information:
Scheduled Time: Wed, Nov 1 - 3:30pm - 4:50pm Building/Room: Convention Center / 404A
Title Displayed in Event Calendar: Evaluating Voice Stress Analysis Software for Deception Detection
Session Participants:
Chair: Christine R. Crossland (National Institute of Justice)
How does VSA work? The Science
Laura J. Pointon (University of Oklahoma), Deidra Upchurch (Oklahoma Criminal Justice Resource Center), Kelly R. Damphousse (University of Oklahoma)
How does VSA work? The Training and the Practice
Deidra Upchurch (Oklahoma Criminal Justice Resource Center), Laura J. Pointon (University of Oklahoma), Kelly R. Damphousse (University of Oklahoma)
Evaluating Voice Stress Analysis Software for Deception Detection
Kelly R. Damphousse (University of Oklahoma), Laura J. Pointon (University of Oklahoma), Deidra Upchurch (Oklahoma Criminal Justice Resource Center)
Discussant: Mike O'Shea (National Institute of Justice)
Abstract:
The session consists of three papers related to an NIJ-funded evaluation two of the more popular VSA tools currently on the market (Layered Voice Analysis and Computer Voice Stress Analyzer). A random sample of arrestees in a county jails were interviewed during the booking process to simulate the stress of a detention-related interview. Each respondent was asked a series of questions about recent drug use. The research team compared the computer analyses of the verbal responses to urinalysis results of urine samples provided by the arrestees.
I'm dismayed that this pseudoscience made it into my professional association's national conference and that the NIJ is involved. Luckily, the session I'm participating in is at a different time. I'll be sure to question all of the presenters on their view of the NAS report and Humble's admission that the CVSA cannot detect lies...
Of course, if they do their evaluation correctly, it should show that the CVSA does not work...