How chatbot communication styles impact citizen reports to police: Testing procedural justice and overaccommodation approaches in a survey experiment

Understanding how chatbot communication styles impact SAR.

Abstract

Objective: We developed and tested a chatbot for reporting information to police. We examined how chatbot communication styles impacted three outcomes: (a) report accuracy, (b) willingness to provide contact information, and (c) user trust in the chatbot system. Hypotheses: In police-citizen interactions, people respond more positively when police officers use a combination of power and solidarity in their communication. We expected that this would hold for citizen-reporting chatbot interactions. Method: We conducted an online survey experiment with 950 U.S. adults who approximated the population on key demographics. Participants watched a video of a suspicious scenario and reported the incident to a chatbot. We manipulated and programmed the communication style of a generative pre-trained transformer chatbot to include elements of the power-solidarity framework from linguistics to create a 2 (power: low vs. high) × 2 (solidarity: low vs. high) design. We then compared three outcomes across conditions. Results: The high power–high solidarity condition yielded the most positive responses. Relative to high power–high solidarity reports, low power–low solidarity reports were less accurate about the individual involved. Trust in the chatbot and willingness to provide contact information did not vary across conditions. Conclusion: Findings contributed to criminological, linguistic, and information technology literatures to show how communication styles impact user responses to and perceptions of a chatbot for reporting to police.

Findings contributed to criminological, linguistic, and information technology literatures to show how communication styles impact user responses to and perceptions of a chatbot for reporting to police.

Full Paper