New method helps identifying murder suspects more quickly

News - 07 October 2021 - Webredactie

Delft PhD student Daan Sutmuller has developed a new, effective method in collaboration with the police that supports homicide investigation teams in collecting and prioritising interesting persons in the investigation more quickly. The promising method should make the investigation work more manageable and ensure that more homicide cases are solved. Sutmuller obtained his doctorate Thursday 7 October 2021 with his dissertation 'Murder investigation in the digital age'.

Complexity of investigation
Although the number of homicides has decreased worldwide in recent decades, the percentage of unsolved homicide cases has increased in the same period. The digitalisation of society has made more and more data available to homicide investigators in order to solve murders. Think, for example, of security cameras that record the street scene and databases with personal data. How do you decide from all this data which persons are interesting in a murder investigation? Little is known about this. Interpreting and weighing pieces of evidence, ranging from eyewitness testimony to highly sophisticated digital and forensic evidence, and arriving at a suspect from there in murder investigations has become increasingly complex due to this enormous bulk of available data. 

Effective method
Sutmuller, who next to his PhD research works as an investigative psychologist with the police, believes there is room for improvement: 'In every investigation, a team has to start again from scratch, for example to determine what a piece of evidence is worth. Every time, the same pros and cons had to be weighed. I wanted to develop an effective method that supports the police in collecting and prioritising persons of interest so that more cases can be solved.' Equally important was that the method reduces tunnel vision in the investigation team and can be used within existing privacy laws. He developed the case specific elements library (C-SEL); a promising method that achieves better results than the currently used methods. 
Library of building blocks
The developed method is actually a kind of library in which pieces of evidence have been given a certain value. These pieces of evidence are building blocks that can be used to determine which persons are interesting to investigate further in a homicide case. The building blocks point the team of investigators in a certain direction to find the perpetrator. The model is transparent, so that at a glance it is clear which choices have been made. This is useful when an investigation is transferred or when it concerns a cold case.

What makes C-SEL different?
The C-SEL method differs from current methods in three ways. The inclusion or exclusion of persons in the investigation is based on 'opportunity and motive' instead of 'opportunity and means'. In addition, a group of 107 experts has assigned a value to various pieces of evidence. The underlying factors of 'relevance' and 'credibility' of a piece of evidence were also taken into account based on a model that was specifically developed to interpret evidence in criminal investigations. 

Positive test results
C-SEL was tested in three real-world Dutch homicide cases involving one perpetrator and one victim. The results are promising. In all three cases, the perpetrator was included in the top 5% of persons of interest in the study. ‘The recorded number of persons of interest is on average lower than in the current methods, which means that the investigation team can work more effectively,' says Sutmuller. The model can be used within the existing privacy legislation and contains properties that reduce the chance of tunnel vision. ‘Of course, it is not the case that we can blindly trust this methodology now. A lot of testing is still needed, for example by using it in solved cases, in cold cases and in ongoing cases. And the model is still to be given a user-friendly interface and must be encapsulated in a software package, but I am very enthusiastic about the initial results.’

/* */