Police Consider the Use of AI to Solve Cold Cases
Some of Britain’s most notorious cold cases could be solved with the help of artificial intelligence after a trial found it can review evidence which would take 81 years of man-hours in less than two days.
A trial by Avon and Somerset Police, using an Australian-developed AI tool called Söze, tested the digital investigation platform on 27 complex cases. The results showed it was able to review all the evidential material in 30 hours. It is estimated it would have taken up to 81 years for a human to manually review the same material.
This is how AI currently contributes to solving police cold cases:
Analysing Large Data Sets
Cold cases often involve vast amounts of data—witness statements, forensic evidence, surveillance footage, and reports. AI excels at processing and analysing large datasets, identifying patterns, and cross-referencing details that may be missed by human investigators due to the sheer volume of information. This could uncover new leads or connections between cases.
Facial Recognition
AI-powered facial recognition can be used to scan photos or videos related to cold cases and identify persons of interest, even if their appearance has changed over time. This can also extend to matching unidentified victims with missing person databases.
DNA Analysis
Advances in DNA sequencing and genetic genealogy powered by AI have already helped solve cold cases. AI algorithms can analyse DNA data faster and more accurately than ever before, finding familial links and identifying unknown individuals through public genetic databases. For instance, the Golden State Killer case (in the USA) was solved using a combination of genetic genealogy and AI-assisted DNA matching.
Natural Language Processing (NLP)
AI’s natural language processing can sift through historical case files, interview transcripts, and witness statements to identify important keywords, patterns, or contradictions. NLP can also compare cases to find commonalities between cold cases that may not have been linked before.
Predictive Policing & Behavioural Analysis
AI models can be used to analyse criminal behaviour, helping law enforcement predict the next steps of a perpetrator or provide insights into their psychological profile. This can be particularly helpful in cold cases involving serial offenders.
Linking Cold Cases
AI can identify links between cold cases that span across jurisdictions, even if the original investigators did not communicate or collaborate. It can also help match unsolved cases with other similar crimes based on geographic location, method of operation, or victim profile.
Social Media and Online Investigations
AI can also comb through social media, forums, and other online platforms to track down new leads or resurface cold cases that have generated public interest. In some cases, this might bring out new witnesses or identify suspects.
Challenges
Despite AI’s potential, there are challenges:
· Data Quality: AI systems need good-quality data, and cold cases often involve incomplete or outdated information.
· Bias: AI models can sometimes reflect the biases present in the data they are trained on, which can lead to skewed or inaccurate results.
· Privacy and Ethical Concerns: Particularly with genetic genealogy and surveillance technologies, there are significant ethical and privacy concerns that need to be navigated carefully.
In short, AI is already helping solve cold cases by enhancing traditional investigative methods, and its role is likely to grow in the future. However, it works best as a tool to assist human detectives, rather than a standalone solution.
As defence lawyers we will ensure that any machine generated evidence is properly tested to ensure it is evidentially sound.
How can we help?
If you need specialist advice in relation to any criminal investigation or prosecution, from the initial investigation through to court proceedings, please get in touch. Call John Howey on 020 7388 1658 or email jhowey@jfhlaw.co.uk. Let us help.
Image credit: “Artificial Intelligence & AI & Machine Learning” by mikemacmarketing is licensed under CC BY 2.0. Image via VPN R Us (now ceased trading).
