Blue Line

News
DEDICATED TO RESEARCH

In the April 2012 edition of Blue Line Magazine a story was posed “Questioning the assumptions – A critique of the Violent Crime Linkage Analysis System (ViCLAS).” Written by researchers Brent Snook, Craig Bennell, Paul J. Taylor, John C. House, Sarah MacDonald and Kirk Luther, its assumptions must not go unchallenged.


May 8, 2012
By Larry Wilson

2006 words – MR

Dedicated to research
Real proof comes from real life with ViCLAS

by Larry Wilson

In the article “Questioning the assumptions – A critique of the Violent Crime Linkage Analysis System (April 2012 ),” the authors argue that “despite its widespread use, ViCLAS hasn’t undergone sufficient scientific testing and consequently may not do what it is supposed to.”

Advertisment

ViCLAS may not have undergone scientific scrutiny to the degree the researchers would prefer but it does what it is mandated to do. They did not clearly articulate specifically what they felt ViCLAS is supposed to do. I assume from the article’s focus they felt its purpose is to link cases. In fact, its actual mandate is to “to encourage and facilitate the exchange of information between law enforcement agencies for the purposes of identifying and tracking serial offenders of interpersonal violence.”

Although ViCLAS does link cases, it fulfills its mandate in several other ways. Two examples:

  • An unidentified female calls a police communications centre from a payphone and said she planned to abduct and rape two children. The agency called the provincial ViCLAS Centre, which worked with the National Sex Offender Registry (NSOR) to identify area sex offenders. One female with a history of offending against young children stood out. The agency arrested her within an hour and eventually charged her with the offence.

  • An investigator contacted a ViCLAS Centre about found human remains which were badly decomposed. Investigators tried other methods and tools to identify the person without success. The specialist had only a physical description and some tattoos to go on but was quickly able to identify a missing person case on ViCLAS where the tattoos matched. Investigators confirmed it was the right person.

ViCLAS does more than linking cases. For example, if police are confronted with a missing or abducted child, ViCLAS would work with the NSOR to conduct “tactical” queries. It can quickly identify sex offenders with a preference for children living in any neighbourhood in Canada. All ViCLAS managers give these cases top priority.

The researchers critique ViCLAS based on four assumptions they say are fundamental to ensuring its effectiveness, concluding each requires further research:

{Assumption 1: Data entered into the system are reliable}

The authors address inter-rater reliability with this assumption. Essentially, it examines how often investigators will agree with what happened when presented with the same data. If there is little agreement then it impacts on the reliability of their responses. The authors use the example of ViCLAS requiring a judgment as to whether a victim was targeted. If two people come to a different conclusion when presented with the same facts, then that response cannot be relied on.

They point to two studies to support their view that there are issues with inter-rater reliability. One was actually led by our office (Martineau and Corey) and the authors conducted the other.

The problem with both is they don’t represent reality. In these artificial studies there is no peril or benefit to those completing the book. They are aware the exercise is for a study, therefore there is no incentive to get the facts straight. In the real world, investigators understand that inaccurate data could impact on the potential to identify linkages between cases.

The candidates also had no first-hand knowledge of the case. Investigators strive to get as much of their case down on paper as possible but many of the details remain in their memory. Using the authors’ example of whether or not a victim was targeted is a good example of something that may be easy for the original investigator to decide without even referencing their notes but may be difficult to conclude based on a written report.

We agree with the authors that inter-rater reliability could be improved but disagree that we need more research in this area. We already understand the value of data reliability and have gone to extraordinary lengths to ensure it is at the highest standard possible. One way we do this is by providing investigators with a thorough explanation of the meaning of individual questions and response options. In addition to numerous training sessions, we have a “field investigator’s guide” available to all police officers that provides a detailed explanation of each question in the book.

We continually examine our data to identify problem areas and address any identified issues. There have been a number of instances where we have formed system review committees for this very purpose, resulting in the complete redesign of the ViCLAS book and the reduction of the number of questions from 263 to 156.

Our most recent system review committee, held in 2009, had the benefit of the results of the inter-rater reliability study conducted by Martineau and Corey. Martineau was present and provided valuable input. The recommendations flowing from that meeting will prompt further changes to the new system when it is released.

{Assumption #2: Data are entered accurately into the system}

This assumption addresses the validity or accuracy of the data. The authors point out they are not aware of any evaluation of the extent to which stored ViCLAS data is valid and quite correctly point out, “…the quality of linkages will only be as good as the accuracy of the data.”

It would be naive to imagine a database with over 400,000 records, each containing hundreds of individual data elements, would be error free, even under the most ideal conditions. It is realistic to assume investigators completing the original submission or data entry clerks entering the books may make the occasional error or omission.

Of one thing we can be sure: where humans are involved, there is the opportunity for mistakes. That’s why we continue exploring ways to minimize the opportunity for errors by using quality control processes before the book makes it onto the system. In most cases, an investigator’s supervisor will review the submission and then send it on to the detachment/department’s ViCLAS coordinator, who reviews it again. If there are no issues, the coordinator will send it on to the ViCLAS Centre where it is processed.

Data can be submitted in paper form or electronically. If it’s on paper, a trained team member does another quality review. This may result in minor changes or the entire submission being returned to the investigator. Once it meets an acceptable level of quality it is entered into the ViCLAS database.

The electronic version of the ViCLAS book has built in quality control filters that have lead to an improvement in submission quality. It also negates the authors’ concern regarding the books not being inputted verbatim into the system by data entry personnel because there is no key stroking involved. The ViCLAS staff imports the book electronically and then performs a quality review. The next version of the ViCLAS booklet will be web-based; investigators will submit all of their cases electronically.

{Assumption #3: Criminal behaviours are consistent and distinct}

The authors’ third assumption “…is that offenders commit their crimes in a way that is distinct from other offenders and consistent enough over time to allow their series to be recognized. This is important because analysts linking crimes appear to use modus operandi behaviours and/or behavioural signatures in addition to information about the offender and victim (Santtila, Pakkanen, Zappala, Bosco, Valkama, & Mokros, 2008)…”

ViCLAS specialists do not make this assumption and have never been taught that an offender’s MO does not change between cases. We go to great lengths to explain that it usually does change and provide a number of possible reasons for the difference in some behaviours across offences committed by the same offender.

The offender’s behaviour is only one aspect of a ViCLAS analysis. There are numerous other considerations outside the context of MO, ritual or signature behaviour that go into the determination a potential linkage exists across offences. These can include, but are not restricted to: offender description, personality, availability and opportunity; geography; crime scene analysis; victim/offender interaction; offender/victim risk; physical evidence (including hold-back) and even intuition.

{Assumption #4: It is possible to identify linked crimes}

The authors say “…people who have received specialized training to link crimes are able to accurately identify serial crime…” and they are not aware of any research examining the impact of linkage training on linkage decision accuracy, nor research examining the performance in the types of linking tasks analysts face in reality.

The only available research, they add, examines how law enforcement personnel (and members of the public) not formally trained in linkage analysis performed on simulated linking tasks and explained such a study using burglary cases.

There is, however, a significant difference in the methodology and degree of training required in linkage analysis between property crimes and crimes of an interpersonal nature. As such, trying to draw any reliable comparison between these studies and how ViCLAS analysis is conducted is a stretch.

While we are not opposed to scientific examination, the fact is, we can produce numerous concrete examples to demonstrate that trained ViCLAS specialists can identify linkages between cases. They prove this each year on the ViCLAS specialist course.

After taking an intensive three week course, candidates are required to conduct a practical linkage analysis exercise. They are given a target case and asked to identify any other cases in the system that may have been committed by the same offender. To pass, they must not only identify the other cases but also explain to a panel of experts their thinking processes used to get those results.

Notwithstanding their success in the classroom, the real proof comes from real life. The following example was reported in our latest newsletter:

Victoria Police arrested the offender of a brutal sexual assault on a sex trade worker, following an 11 month investigation in which DNA evidence linked him to a 2009 unsolved attack on another Victoria sex trade worker. In February 2011, a sex trade worker reported a violent sexual assault to Victoria Police and gave them a detailed description of the offender.

The victim was able to get the offender’s license plate number. It belonged to a female whose boyfriend matched the offender’s description; however, the victim was unable to identify this male in a photo lineup. In May 2011, a ViCLAS analysis was conducted on the Victoria file and a potential linkage was made to another unsolved violent attack on another Victoria sex trade worker from 2009. A potential linkage report was completed and sent to Victoria PD. The investigators followed up on the files, including forwarding exhibits to the lab.

Male DNA profiles were located on exhibits from each case. These unknown DNA profiles were sent to the Crime Scene Index. The offender was on the convicted offender DNA registry. In January 2012, the two cases were confirmed to be linked with the offender identified as the original suspect.

{Conclusion}

The RCMP’s Behavioural Sciences Branch has a position dedicated to research and ongoing research projects, some in partnership with universities and other law enforcement agencies. We recognize the ViCLAS database is the largest source of data on crime of an interpersonal nature anywhere and appreciate its research potential.

Nonetheless, we must be extremely careful and selective when allowing ViCLAS to be used for research purposes, particularly when it involves researchers from outside law enforcement. The data includes highly sensitive information on both the offender and victim which we are compelled by law to protect and ensure it’s not used other than for the purpose it was gathered. The goal of the research must contribute to the advancement of law enforcement in some way.

The authors made some valid points but may have left readers with the impression ViCLAS does not work and investigators may be wasting their time submitting cases. I want to assure your readers that it does work and investigators are not wasting their time.

bio box

RCMP Inspector Larry Wilson is the officer in charge of the federally managed Violent Crime Linkage Analysis System in Ottawa. Contact him at 613 843-5999 for more information.