Artificial Intelligence will be a major game changer
Artificial Intelligence (AI) is starting to change the face of police investigations in an unprecedented way. It can help triage or filter material, locate where to look in a caseload or highlight material that is likely to be child sexual abuse material. As a result the technology was reported as one of the most helpful technologies to aid child sexual abuse investigation in the NetClean Report 2019.
By Jim Cole, Supervisory Special Agent, Homeland Security Investigations (USA), Special Agent in Charge Nashville, TN
The technological development that is having the single most impact on child sexual abuse investigations is Artificial Intelligence (AI). I have used AI for many years now and in the past two or three years there have been some incredible developments.
The future of AI
The future of AI is exciting, and we are still only in the beginning of where AI can take us, but it is clear that visual analytics will change beyond our imagination.
An example of current really interesting work is a research project that aims to marry the human brain’s neural networks with AI neural networks. The human brain has an amazing capability to understand context, something machines are terrible at. In this project images are shown to a person (police officer) at a faster pace than the brain can cognitively register. The brain still reacts to the content of the image, and violent content elicits a heightened stimulus. By reading a brain’s EEG, AI can use those reactions to group images together. In the future this could be used during the first sift through of case data, to help law enforcement go through material faster.
There is also work on AI being used for text analysis to detect certain text sequences in chats and messaging apps to push content for review. That could be used in both child sexual exploitation and terrorism investigations.
Another future use is resource intensive undercover operations where a detective chats with suspects. With some human monitoring, that could probably be AI driven instead. Or when doing a house search, images from the premises could automatically be matched to see if the location matches unsolved child sexual abuse series that we have.
These are just some examples; I think that we will see AI move into areas that we are far from contemplating now.
We are still some way from the future I’ve painted above, but AI can already be very useful. It is already a good indicator and provider of investigative leads.
Instead of just looking at a vast pool of data in a case, image classifiers can help us group together images with visual similarity and highlight images that might be of particular interest. This makes the data both easier and faster to review, making investigations more efficient.
“By reading a brain’s EEG, AI can use those reactions to group images together. In the future this could be used during the first sift through of case data, to help law enforcement go through material faster.”
Expectations vs capability
The challenges that we face with AI right now, is the general poor knowledge level and people’s expectations of it. In the general police community there is a lack of understanding of how AI works, and how the algorithms “learn” and there is a tendency for investigators to rely too heavily on the result from the computer.
As a result, expectations are exceeding the capabilities of the technology. There is a belief that AI classifiers will find all child sexual abuse material in a case, and that it will find everything correctly – which it doesn’t. It is not a result that can be trusted without review or that can be used in court, it still needs to be manually validated and investigated.
Therefore, we need to make sure that we train people in the police community to understand the strengths and limitations of AI, so that they use its potential efficiently and correctly.
Automate manual work
I don’t think there will be a time during my career nor lifetime where we can blindly rely on computers and algorithms. However, in the not too distant future we might be able to automate up to ninety percent of what we today are doing manually. This will cut down on the time and resources that we have to spend reviewing images and will free up time for people to work on more new cases.