KINtalk: Impact of artificial intelligence on our work: predictive policing

Reading Time: 4 minutes

On 11 October, 2019 at KIN Center for Digital Innovation, we talked about the impact of artificial intelligence (AI) on work, delving into the case of the Dutch police and predictive policing.

Professor Marleen Huysman opened the floor by comparing different perspectives on the study of AI. She discussed the labor perspective which focuses on macro-level changes on the whole labor market; the critical perspective which discusses ethical issues around AI in a  theoretical/conceptual manner; the business perspective which talks about the potential of AI in tackling various challenges around organisation and innovation. Then Marleen introduced the practice perspective – a stance taken by the researchers at KIN – which looks into the actual practices around the design, control, and use of AI with the aim of going beyond the hype and working together with practitioners.

After the brief yet comprehensive overview of different vantage points to research on AI and work, the floor was given to Dick Willems, a data scientist at the Dutch police and creator of the Dutch predictive policing algorithm: ‘Crime Anticipation System (CAS)’.

As any other organisation with finite resources, the Dutch police were faced with the challenge of: What should we do with our (limited) manpower? Where should we allocate our capacity? The idea of identifying ‘hotspots’ – classifying areas according to the frequency of crimes – was in fact not a new practice in the police. However, they were still faced with the problem of police officers reading these ‘heat maps’ in a subjective manner!

So, the Dutch police came up with an objective way of generating these heat maps, using two types of data. The first type of data is location-specific and relatively static, including information around: How old are the people living in this area? What is the average income of people in this area? The second set of data regards the actual crime history and is more fluctuating, thus requiring frequent updates. This can include information such as the number of crimes during the last week in a given area.

Simply put, CAS was created in 2012, with an organizational aim to make optimal use of the manpower to tackle crime in the Dutch police by generating heat maps which signal priority areas of manpower allocation. Dick explained that they tried various classification techniques and decided to use logistic regression, partly due to the fact that the model needs frequent updates of crime history data to stay relevant. Then Dick went on to show an example of a heat map with a “good performance” where 3% of the map was coloured as hotspots among which 20% was predicted correctly and 32% missed, but still considered close enough.

Needless to say, the development of CAS was not the end of the story. Getting officers to actually use this technology that they had never seen before was another challenge. Dick explained the hardship in these words: “If you give these maps to police officers, they will not use them. They don’t want to change their work practices. That’s not in their personalities.”

For the police officers to actually use the computed output, they needed someone that would work as a middleman between the computed output and the police officers. Hence, the creation of ‘information specialists’ who are in charge of fusing the quantitative information from CAS with the qualitative information from the community police officers, for example. In the end, Dick summarised that the whole information system had to change and this did not happen just by introducing the technology, but by figuring out a way to work with it on an organizational level!

As the third speaker of the session, Lauren Waardenburg, a PhD candidate at KIN who has studied the Dutch police around CAS for the past three years took the stage. The core question in her presentation was: How is predictive policing actually used in practice? – very much in line with the practice perspective of KIN. She first introduced the three common assumptions about AI: (1) The impact of AI on individuals; (2) Job loss and deskilling; and (3) Efficiency and objectivity. Then with her empirical data, she basically debunked them one by one!

First off, the impact of AI is found not so much in an individual that loses his job, but more in a network of individuals that are mobilised to enable the implementation of a technology. Secondly, despite our liking for doom stories of deskilling, the reality around AI presents us more often with stories of re-skilling. Lauren says that ‘information specialists’ are an excellent example of that other side of the story that we seldom talk about. Another aspect of information specialists leads to the third point that technology does not automatically lead to more efficiency and objectivity. 

Has police work become more efficient? Well, over time she has observed police officers spending more and more hours behind their computers, putting what they have seen and heard and done into CAS. Then, has police work become more objective? Well, the police still have information specialists that need to fuse the ‘objective’ data with their ‘subjective’ insights. This links back to the point made earlier by Dick Willems that introducing technology alone does no magic and that we need a careful orchestration of resources and practices on an organizational level.

After Lauren’s presentation that nicely summarised the Dutch police case and invited the crowd to take a conceptual leap to ponder about technology, work, and organisation in general, the floor was open for Q&A. Several rounds of interesting questions and answers followed including questions such as: Is the new technology embraced by police officers or do they resist? What do they think of the changing nature of their work? Is there any element of Artificial Neural Network used in CAS? We had a very interesting discussion and many people stayed around for drinks, snacks, and networking!

Interested in what would be discussed at our next KINtalk? Stay tuned for more real and tangible stories around cutting-edge digital technologies, innovation, and the practice story of it! We hope to see you at our next KINtalk!

Author: Bomi Kim