Explanations Increase Citizen Trust in Police Algorithmic Recommender Systems: Findings from Two Experimental Tests
Esther Nieuwenhuizen, Albert Jacob Meijer, Floris Bex, Stephan Grimmelikhuijsen · Public Performance & Management Review · 2024
A long-standing question in e-government research is how to maintain citizen trust in digital encounters with the government. This question is even more pertinent as algorithmic recommender systems (such as chatbots) are now becoming an integral part of digital public service delivery. The literature suggests that the explanations that these systems provide for their recommendations are crucial to maintaining citizen trust in digital encounters, but so far the empirical research into this relationship is limited. To test the effects of various explanations provided by algorithmic recommender systems on citizen trust, we conducted two experimental studies. We developed a mock version of an actual algorithmic recommender system used by the Dutch police and tested it in two representative survey experiments. Study 1 (n = 717) tested the effects of procedural, rationale and combined explanations. We found that providing any explanation increased trust and made citizens more likely to follow an algorithmic recommendation. Study 2 (n = 1005) investigated whether providing a directive explanation—specific instructions for achieving a desired service outcome—increases trust, building a more nuanced understanding of the relationship between explanations and trust in algorithmic recommendations. We conclude that explaining algorithmic recommendations—in any form—strengthens trusting beliefs, trusting intentions and trust-related behavior in citizens receiving digital public services. This may suggest that trust in algorithmic recommendations increases when citizens see that governments make an effort to provide an explanation, regardless of the nature of this explanation.