News
Increasing trust in the police: research contributes to National Police’s Intelligent Crime Reporting tool
Recently, this research project has directly contributed to fine-tuning the use of the National Police’s Intelligent Crime Reporting (ICR) tool. This direct application of the research findings can contribute to strengthening citizens’ trust in the use of algorithms by the police.
Explaining algorithmic decisions
In our first study, conducted within the ALGOPOL project, we aimed at determining the effect of algorithmic explainability on citizen trust. How can the police strengthen citizens’ trust in its Intelligent Crime Reporting (ICR) tool? PhD candidate Esther Nieuwenhuizen, together with the Utrecht Policelab AI, examined this question. Dr. Stephan Grimmelikhuijsen and prof. dr. Floris Bex supervised the study.
Intelligent crime reporting
The intelligent crime reporting-tool is a smart decision-making system used by the Dutch police. The system uses algorithms to assess reports of online trade fraud, such as fake web shops. We tested how different types of explanations effected citizen trust in the decisions made by the ICR-tool, by means of an experiment.
Providing specific motivations substantially increases trust
Esther Nieuwenhuizen: “The results of our research show that a justifications explanation – a motivation for the specific decision of the ICR-tool – had a substantive positive effect on trust. A procedural explanation – information about how the algorithm came to its decision – only had a small positive effect.” Based on these findings, the police asked Esther to put the recommendations into practice.
Making impact at the police
Esther rewrote all decisions of the ICR-tool in which citizens are advised not to report their case. In each case, she incorporated a justifications explanation to show the user why it is not a case of online fraud. For instance, because the web shop they reported is listed as “trustworthy” in the police system, or because the person did not wait long enough before a report can be filed.
Before the researcher made these changes (with the previous original explanations given by the ICR-tool), 72% of the people who received the advice to not report the crime, still chose to file a report. Providing a justifications explanation can reduce this percentage to 9.4%, as shown in the experiment. Nieuwenhuizen: “In the coming months, we will learn whether the outcomes in practice are as substantive as the results of our experimental study. Hopefully, rewriting the explanations in the decision-making system will help in building trust among users of the ICR-tool.”
Maintaining the trust of citizens
This study provided valuable new insights into how trust in algorithmic policing can be strengthened. Considering the increasing use of algorithms by the police, the ALGOPOL research team will further research the claim that value-sensitive and transparent algorithmization are needed to ensure that public organizations -such as the police- maintain the trust of citizens.