Research of Martijn Wessels shows how police officers can deal with artificial intelligence within the polcie organisation. As a case study he examined the predictive policing system CAS. Below is a summary of the results and full report to read or download.
Law enforcement agencies around the globe are turning to algorithms that help to make spatiotemporal risk analysis of where and when criminal activity is most likely to occur. This so-called predictive policing is facing a lot of scepticism and criticism. It is being disputed whether this type of policing is actually increasing the effectiveness and efficiency of the police, as real empirical evidence is scarce or hard to isolate to determine the effectiveness of predictive policing. Additionally, there are also voices of concern which argue that these systems are lacking transparency and might lead to profiling and stigmatisation of societal groups. However, the gap that is recognised in this debate is that it is still relatively uncovered how professionals actually use this system, whilst these insights can help to put the discussion into context. This is a crucial factor, as human agency ultimately determines to what extent unintended consequences come to fruition, it is crucial in the evaluation of this type of policing.
In order to shed light in this theoretical and empirical gap, the Dutch policing context has been studied. As the Dutch National Police is already using an equivalent of predictive policing software (the Crime Anticipation System, CAS) for a longer period of time, this research site is chosen. In this specific context, it turns out to be that the intelligence specialists are the group within the police organisation that have the purpose to construe advice for the operational level of the police organisation, in which they can include CAS. As a consequence, this group is being studied to understand how they use this system in their practices, and to see how organisational structures influence this utilisation. The central research question of this thesis reads:
How and to what extent do intelligence specialists of the Dutch National Police utilise the Crime Anticipation System in their practices and thereby interact with existing organisational structures?
Orlikowski’s (2000) theory of technologies-in-practice is administered to help understand how technology is being used in an organisational context. This practice lens argues that the way how technology is utilised by professionals is influenced by enacted organisational structures, and in turn simultaneously affects these structures. To fit the purpose of this research, Orlikowski’s (2000) model is adapted to help study the utilisation of algorithms, as this type of technology also requires an assessment of the output: the expectations of CAS need to be trusted to some extent in order for the intelligence specialists to include this system in their practices. Hence, this understanding is added to the analytical model that has been used which leads to an adapted analytical lens: the algorithm in practice lens.
Through a combination of semi-structured interviews and the analysis of policy documents, it is uncovered how intelligence specialists utilise CAS in their practices and how this is being influenced by enacted organisational structures. The first structure recognised is the ongoing trend towards intelligence-led policing within the Dutch National Police. Hereby, the police aims to give data and information a prominent role, as their organisational structure are altered to stimulate this information exchange. As a result, the tasks of the intelligence specialists are now becoming more standardised, and pose a central role in the police organisation. Due to this centrality, the intelligence specialists are deemed to provide all the relevant information in a suiting manner to the operational level of the police organisation.
As a consequence, they have extensive contact with the operational level to make sure that their advices match the needs of the operational level in a suiting manner. Hence, this is the second organisational structure that influences the practices of the intelligence specialists: the needs of the operational level. As a result, this structure and the agency of intelligence specialists are in constant interaction.
The norms of the intelligence specialists, result of the intelligence-led policing trend and the interaction with the operational level, influence how CAS is being used. The norms of the intelligence specialists are recognised as the provision of advice that has the highest quality (i.e. usefulness). As a result, the interpretative schemes of the intelligence specialists in respect to CAS are of vital importance to how it is being used. If CAS is not deemed as added value to their advice, it will not be utilised. These interpretative schemes are (partly) influenced by the last organisational structure: the opinions of the direct social environment of the intelligence specialist. As several intelligence specialists learned of CAS via peers, the attitudes of their colleagues are important: if they are sceptical, it is possibly more likely that the new intelligence specialists will start sharing this perception.
In this study, three categories of CAS usages are recognised:
1) The supporting colleague: CAS is used as an additional element to their advice. Hereby, the intelligence specialists see benefits of using CAS as it helps to prevent tunnel vision, ‘thinks’ differently as it combines multiple sources of data, is rapid in analysing large amounts of data, and helps to provide advice when the intelligence specialist do not have enough information on its own. Nevertheless, they still deem their own expectations and experiences as most important, and provide advice on the local trends and issues that are observed.
2) The unsuitable or unnecessary colleague: CAS adds no value to the advice of the intelligence specialists as it either is inaccurate in its predictions, or has no benefits to the operational level. As a result, the intelligence specialists develop advice based on the their own expectations and knowledge.
3) The directive colleague: CAS is used as a directive tool to directly advice the operational level. The intelligence specialists explicitly communicates the output of CAS and perceives this as a legitimate tool for decision-making.
These different usages of CAS have consequences for the intelligence-led policing trend. In the first category, the advices are derived on a combination of own experiences, knowledge and observations, and the statistical information from CAS. Hence, this pushes the notion of intelligence-led policing, as explicit information is having a prominent role. This is pushed even further when CAS is being used as a directive tool, where the own experiences are seeming to become less prominent. The intelligence specialists that do not use CAS base their advice on their own presumptions and knowledge, and therefore the intelligence-led policing trend is more reliant on the (implicit) assumptions of the individual intelligence specialists.