The second season of the television programme Hunted opens on Monday 6 November (AvroTros, NPO3). In this programme, 14 ‘fugitives’ spend 21 days trying to escape capture by a team of expert professional investigators. This season the ‘hunters’ will be supported by QUIN, a predictive model for fugitives developed by TNO.
QUIN – short for ‘Question & Investigate’ – is the brainchild of Selmar Smit, an artificial intelligence researcher. The software was developed in collaboration with the Dutch police, with the aim of reducing the analytical workload in liquidation cases. “We chose the name QUIN for a reason. The software works very like the character of Mr. Quin in the Agatha Christie books, a mystical figure who comes and goes and who whispers ‘Have you thought about that?’ into the chief inspector’s ear.
Our QUIN is a system based on the fact that a crime can only be carried out in a finite number of ways. Every crime resembles another. That means that you can guess what a suspect might do. QUIN can help analysts to process the available data and to predict a suspect’s next moves,” Smit explains.
“On the basis of the information known about a current case and the same information drawn from old cases, QUIN can make predictions”
How does it work?
On the basis of the information known about a current case and the same information drawn from old cases, QUIN can make predictions on such things as personal details, case-related factors such as times, locations and transport choices, and the residential locations of family and friends. Smit: “QUIN calculates the distance between two cases and the model compares a variety of aspects that appear similar. It can then make predictions about where a suspect might be located, or might sleep – in a hotel, for instance, or with friends, or somewhere else – and estimates the likelihood that the suspect is still at a known location.”
Hunted as validation
Validating the effectiveness of the tool Smit developed is difficult, however, without access to the sensitive information surrounding ongoing cases. “Logically, you don’t get access to police data if you have no indication of your effectiveness – and you don’t want to disrupt ongoing cases with possibly faulty information. So taking part in the Hunted programme looked like a great way to break out of this vicious circle. There’s no question of sensitive information, but the scenarios are reasonably realistic. We got in touch with the producers, and luckily they were enthusiastic about the idea. We made QUIN available to the team, and they added me to the team as the analyst. It was a lot of fun to spend three and a half weeks working in the investigation team. An enjoyable experience – and of course, very different from my day to day work,” Smit explains.
“During shoots for the new season we saw that QUIN made quite a few good predictions. It looks like the tool really works!”
To prepare for the task, all data from earlier seasons of Hunted, including foreign versions, were collected and entered into QUIN to learn how these ‘fugitives’ – not genuine fugitives, obviously – had behaved. A new, slick interface was also added to the program. “During shoots for the new season we saw that QUIN made quite a few good predictions,” says Smit. “It looks like the tool really works!”
Developing QUIN was, naturally, a team effort. “I’m the one you see on TV, but there’s a whole team behind me that developed and built QUIN.”
No use to criminals
Smit has thought about whether it was a sensible step to show QUIN on television. Might it not actually make criminals smarter? “It doesn’t make any difference, really. Criminals can’t do anything with it. It won’t have a negative effect on the catchability of criminals. All they can do is to change their behaviour – but then their movements are unnatural, they make mistakes, and that might actually make them easier to catch.”
“This is a run-up to a fundamentally different way of working in the area of data analysis and prediction.”
The police are interested in an ongoing collaboration, and a QUIN pilot may soon be carried out. “Software like QUIN helps the police do what they do best: catch criminals. And this tool could also be interesting for other security organizations and services, too,” adds Smit.
“The work of security services is going to be increasingly driven by information and technology in the years to come,” agrees Krishna Taneja, director of National Security at TNO. “Online, and through devices like smartphones and cars, more data is being collected than the human brain can process and analyse. TNO is looking for solutions to this problem, in close cooperation with the security services and the business community. QUIN is an example of this, and has shown that even in its test phase it could turn raw data into predictions. This is a run-up to a fundamentally different way of working in the area of data analysis and prediction.”