Apps for cops: a realist evaluation of the impact of mobile information technology on police decision-making and on police-citizen interactions

Police services have increasingly embraced digitalization in recent years, and the offered technologies have become more sophisticated. The digitalization of the police is a frequently discussed topic in academic literature as well. While there is already a considerable amount of research on the impact of various technologies, this primarily focuses on their impact in terms of performance (e.g. effectiveness or efficiency). Much less is known about the impact of technology usage on the daily decisions made by police officers and how this translates into their daily interactions with citizens. This research aims to address that lacuna by analyzing the use and impact of one specific technology, namely ‘mobile information technology’ (including the ‘Focus app’ used by the Belgian police). Mobile information technology (hereafter MIT) enables police officers to access a wide range of information (e.g. databases, hotspot lists) and to fulfill reporting requirements (e.g. generating police reports) directly in the field, all using a mobile device (such as a smartphone, tablet, or laptop). Specifically, three research questions were formulated, referring to successive steps in a chain:

To what extent do police officers use MIT?

If police officers use MIT, what is the impact of that usage on police decision-making?

If the use of MIT has an impact on police decision-making, how does that impact translate in the interaction between police officers and citizens?

Previous empirical research on the use of technologies, specifically MIT, has yielded conflicting findings regarding its impact. For instance, various studies have observed that MIT leads to time savings for police officers (e.g. Greenberg, 2017), while this was not confirmed in other studies (e.g. Carter et al., 2016). A clear explanation for these contradictions has largely remained absent. What sets this research apart is its focus on explaining such discrepancies, using the research strategy of 'realist evaluation' (Pawson & Tilley, 1997). More specifically, the study zooms in on differences in the context in which MIT is implemented ('contextual factors') and variations in the configuration and implementation of MIT ('intervention modalities') that could explain differences in use and impact ('outcomes'). All these components are brought together in 'context-intervention-mechanism-outcome (CIMO) configurations. This leads to a fourth research question:

What role do contextual factors and intervention modalities play in the usage and impact of MIT?

To answer the research questions, several hypotheses (theoretical CIMO configurations) were first formulated based on academic literature. These theoretical CIMO configurations were then tested and refined multiple times in the empirical research. Various forms of empirical data were collected in three local police forces in Belgium. Belgian local police forces were a good choice since the national rollout of a new MIT, namely the 'Focus app,' had just begun at the start of the doctoral research. The CIMO configurations were refined a first time based on qualitative data collected in two pilot case studies (two local police forces that were not included as cases in the actual data collection). With this knowledge, an online survey was developed and administered to the chiefs of the local police forces in Flanders and Brussels to select the three actual cases. Both in the two pilot cases and the three actual cases, the data collection methods consisted of a combination of interviews (11 in the pilot cases and 39 in the actual cases) and field observations (11 shifts in the pilot cases and 86 shifts, totaling 688 hours, in the actual cases). Specifically, respondents from the traffic, community and patrol departments were involved in the three actual cases. Case 2 also included the police forces’ 'project teams', which work primarily proactively to address various forms of disturbances. In total, ten services were thus involved. A within-case and across-case data analysis approach was performed, comparing the police forces, the selected departments, and, where relevant, the individual police officers. This allowed for an assessment of the role of variable contextual factors and intervention modalities.

Regarding the first research question (which focuses on the extent to which police officers use MIT), we can broadly conclude that MIT was frequently used in all included departments. However, in most departments, police officers still relied regularly on the alternatives that existed before MIT was introduced. For example, in most departments, database inquiries were still conducted through the control room, and reporting requirements were often (in some departments, still as a standard practice) fulfilled at the office. The thesis provides a detailed description of the circumstances in which police officers chose to use or not use MIT. Various contextual factors, intervention modalities, and mechanisms were identified to help explain those choices. For instance, MIT was only used for reporting in departments where reporting through MIT could be easily handled with standard texts (intervention modality) or in services where on-site reporting was mandatory (contextual factor).

To address the second research question (aimed at examining the impact of MIT usage on police decision-making), three aspects were considered. First, it was investigated to what extent MIT distracts police officers from what is happening around them. A distinction was made between distraction during patrolling and distraction during actual interventions. Concerning the first aspect, it was found that police officers frequently accessed the Focus-app during patrolling, potentially causing them to miss events happening around them. The frequency of this distraction varied significantly between departments. For example, the Focus-app was used more frequently in services where the device was mounted at the front of the vehicle. Regarding the second aspect, it was observed that in most departments, police officers often used MIT during interactions with citizens, which did indeed lead to distraction. However, police officers typically worked in pairs, and only one of them used MIT, allowing the other officer to focus on the situation and the citizen.

Second, it was expected that the capabilities provided by MIT would motivate police officers to take certain actions more frequently. This expansion of actions is referred to as ‘netwidening’, which means broadening the scope of an individual police officer's surveillance net. Specifically, it was examined whether netwidening could be observed with regard to four decision points: the decision to 1) conduct a database inquiry, 2) perform a police stop, 3) engage in proactive patrolling, and 4) issuing official police reports for an offense. Concerning the first two decision points, it was found that, except for three departments (namely the community officers in cases 1 and 2 and the patrol department in case 3), police officers indeed conducted more database inquiries and person stops using MIT. Generally speaking, MIT made police officers less hesitant to conduct inquiries (and person stops) compared to when they had to burden the control room. The ease of using MIT also seemed to lead to reduced reflection on the necessity of the inquiry. Inquiries were often conducted ‘just to be sure’ or to make an initial assessment of whether a police stop would be worthwhile. Regarding vehicle stops, the ability to easily look up license plates (and then proceed to check the owner and the likely driver) before conducting a stop, actually led to more reflection on the necessity of a physical stop. The information from the inquiry was used to assess whether a physical stop was desirable, resulting in netwidening for inquiries but not for the actual interaction with the citizen. Another finding related to police stops was that in cases 1 and 2, MIT led to fewer registrations of these stops. Before the introduction of MIT, every interaction was recorded as soon as a database inquiry was requested through the radio channel. With field officers now able to do their inquiries themselves, registrations only occurred when the police officer deemed it necessary. The threshold for when registration was needed varied among respondents. In case 3, proactive stops were not always registered even before the introduction of MIT. Concerning the third decision point, the decision of where to patrol, it was expected that MIT would be used by supervisors to direct patrols to specific locations. In all three actual cases, MIT was indeed used for patrol assignments. In some departments, the assigned tasks were carried out in addition to the officers' regular patrols, resulting in netwidening, as more patrols were conducted than before the introduction of MIT. In other departments, patrols were limited to the assigned tasks, resulting in geographical netnarrowing. The time spent on patrols in those departments remained the same but was mostly limited to the locations on the assignment list. Regarding the decision to issue official reports for offenses, the ease of doing so in the field with MIT led some respondents to state that they did so more frequently. Others reported that MIT had no impact on their decision to issue official reports. Interestingly, the six respondents who indicated that MIT led to netwidening came from six different services across the three cases, indicating that no relevant contextual factors at the force or department level could be identified. Those who cited administrative burdens as a factor in their decision to issue official reports confirmed this form of netwidening. In summary, the assumption regarding netwidening was partially confirmed but not for all examined decision points.

Third, it was expected that, as a result of their autonomous access to database information, police officers would base their decisions more on 'bureaucratic' database information and less on 'incidental' information (visible elements of the situation, e.g. behaviors, physical characteristics). This type of decision-making, based on knowledge of previously observed criminal or suspicious behavior, is referred to as 'bureaucratic suspicion' (Matza, 1969). For each observed decision point, it was examined whether there was (partially) bureaucratic suspicion and whether it was a result of MIT. The latter was deemed not to be the case if the police officer could have obtained the bureaucratic information consulted via MIT through the control room as well. For the decisions to conduct a police stop, register a police stop, search a person, and search a vehicle, it was indeed found that as a result of the availability of MIT, these decisions were more frequently (partially) based on information from databases than before. Thus, for these types of decisions, there does appear to be an increase in bureaucratic suspicion. The decision to conduct a drug test was also often (partially) based on bureaucratic information, but in most cases, the police officer would have obtained that information through the control room as well. Further, the decision to issue an official report for an offense and the decision to conduct an alcohol test were primarily based on incidental suspicion in most cases.

To answer the third research question regarding the impact on police-citizen interactions, seven dimensions of interaction were identified, inspired by Della Porta and Fillieule (2004), which were expected to be influenced by the use of MIT, namely: (1) increased likelihood of interaction - decreased likelihood of interaction, (2) harsh - soft, (3) reactive - interventionist, (4) little communication - much communication, (5) strict - flexible, (6) fast - slow, and (7) cautious - incautious. Diverse effects were observed on each of these dimensions depending on how MIT was used.

This research contributes to the academic literature by providing empirical support for two important assumptions regarding the potential impact of police digitization. Both the assumption that MIT would lead to netwidening and the assumption that it would lead to bureaucratic suspicion were confirmed but also nuanced by the data. In this way, the research results in an empirically grounded theoretical framework that provides insight into the impact of MIT in different circumstances. This theoretical framework is also relevant to practice since the final, empirically grounded, CIMO configurations allow policymakers to consider which circumstances they should take into account when aiming to achieve a particular outcome (or not). In addition to these specific recommendations, the research also yields some important findings that call for reflection. For instance, the finding that MIT leads to less information being recorded in formal databases is not without risk. It was observed, that information is either not recorded or recorded in more informal databases created within MIT (e.g. in briefing items). This can result in information being lost, and it may also lead to information overload for field officers, making it impossible for them to review all the information. The non-registration of proactive police stops means that neither supervisors nor colleagues are aware that a stop has taken place. It also implies that police officers no longer offer the written justification they typically had to provide for the stop. Moreover, the observation that the additional information provided through MIT is only rarely consulted during an inquiry, raises questions about the necessity of maintaining and offering such a large amount of information. If police officers often do not have the time to search for information in these modules themselves, perhaps the most relevant information should be pushed to them. Given the finding that the ease of using MIT reduces reflection on the necessity and legality of certain actions (e.g. database inquiries or police stops), it is important to encourage police officers to keep reflecting sufficiently about these actions. Oversight mechanisms can provide a potential solution, but they should not discourage the proper use of MIT. What seems more important is a supportive approach that teaches police officers to handle this increased autonomy.

Finally, the observation that police officers appear to base many decisions on database information also calls for reflection. This confirms the need for high-quality, up-to-date databases. Moreover, the Belgian legal framework currently mainly refers to incidental grounds for many actions. For instance, administering a drug test must be justified by reference to visible signs of drug use. 'Being known in the police databases' is not included as a criterion to justify a drug test. Similarly, for other police powers (e.g. searching a person or a vehicle), being known in the police databases can generally only be considered a legal reason in combination with incidental elements that justify the search for information about someone's criminal history. Police officers seem to be aware of these limitations, as they appeared to omit the bureaucratic elements on which their decisions were based in their written justifications. The growing role of information from databases also raises normative questions. There is a risk not only of overpolicing based on visible elements and thus incidental suspicion (which currently receives much attention, particularly in the debates about ethnic profiling) but also based on registered data from the past and thus bureaucratic suspicion.