Feb 17 2012 A national computer system designed by the RCMP to help investigators identify links between violent and serious crimes - and stop serial killers and predators in their tracks - has come under scrutiny by a team of Canadian academics, who say the system's effectiveness has not been rigorously tested despite being in use for almost two decades.
The Violent Crime Linkage Analysis System or ViCLAS, which costs $12 million to $15 million annually to run, contains details on more than 400,000 cases and has been credited in recent months with helping investigators link - and in some cases solve - a variety of crimes, including sexual assaults and the luring of minors on the Internet.
The system has been licensed to police forces in other countries, including Britain, France, Germany and New Zealand.
"People beat a path to our door," said RCMP Insp. Larry Wilson, the officer in charge of ViCLAS in Ottawa. "There's no other system like it."
But in an upcoming article in the journal Criminal Justice and Behaviour, researchers write that anecdotal successes "do not constitute strong evidence" the system works, and that empirical testing of the way data are entered into the sys-tem and analyzed is overdue.
"The jury is still out in terms of whether there's evidence to support its use or not," lead author Craig Bennell, a Carleton University psychology professor, said in an interview. "We need to explore this more before too much confidence is put in these systems."
The need for a crime-linkage database was first identified in the 1980s after the series of child-sex killings by Clifford Olson in B.C. The urgency grew in the 1990s, when a report from Ontario Justice Archie Campbell noted "systemic weaknesses" in the investigation of serial rapist and killer Paul Bernardo in that province.
Anytime police investigate a homicide, sexual assault, child luring or suspicious missing-person or human-remains case, they are expected to enter information about that case into the database. The way they do that is by answering a standard ViCLAS questionnaire containing 156 questions designed to tease out details about the offence, crime scene, perpetrator, victim, weapons and vehicles.
Analysts in provincial ViCLAS centres scrutinize the data and query the system to see if they can find similarities - in physical evidence or behaviour - between different crimes.
When analysts believe they have a match, they will write a report and submit it to investigators. On average, ViCLAS analysts send 175 to 200 "potential linkage reports" each year, Wilson said.
A recent ViCLAS newsletter credited such a report with helping police nab a suspect in two sexual assaults in Golden. A woman was walking along a trail when a man approached her and asked if the cellphone he was holding belonged to her. The man grabbed the woman and began choking her and pulling her pants down. The woman was able to fight him off.
A suspect was identified, but he denied any involvement. A ViCLAS analyst looked at the case and found a similar case from a year earlier in the same community. A woman was walking her dogs when a man approached her, threw her to the ground and pulled her pants down before running off.
The analyst sent a report to investigators and when investigators confronted the suspect again he admitted involvement in both crimes.
Officials concede, however, that ViCLAS analysts currently have no way of knowing what their success rate is because there is no requirement for investigators to get back to them to tell them if their suspicions were correct.
Bennell and his research colleagues say that's not the only problem with the system.
They write in their upcoming article that ViCLAS relies on several assumptions, none of which are supported by empirical evidence.
One assumption is that data entered into the system is reliable and accurate. Bennell's colleague, Brent Snook, a psychology professor at Memorial University in Newfoundland, performed a study in which he asked 10 police officers to review the same case file and then fill out a ViCLAS questionnaire.
Officers' answers matched only 31 per cent of the time, indicating "low levels of reliability."
Snook said it's possible the officers might have achieved better results under real-world conditions. But he said the results could also indicate flaws in the questionnaire's design.
Another assumption that ViCLAS makes is that serial offenders exhibit the same behaviours from one crime to the next. But the authors write that a majority of research has shown only "low to moderate" levels of consistency in the modus operandi of violent offenders.
Finally, the authors point out there has been little research to show that analysts who receive special training to link crimes actually make linking decisions accurately.