Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Are We Asking the Right Questions?: Designing for Community Stakeholders' Interactions with AI in Policing

MD Romael Haque, Devansh Saxena, Katy Weathington, Joseph Chudzik, Shion Guha

TL;DR

Surprisingly, it was found that domain experts (LEAs) often exhibited anchoring bias, readily accepting and engaging with the first crime map presented to them, and community members and technical experts were more inclined to engage with the tool, adjust controls, and generate different maps.

Abstract

Research into recidivism risk prediction in the criminal legal system has garnered significant attention from HCI, critical algorithm studies, and the emerging field of human-AI decision-making. This study focuses on algorithmic crime mapping, a prevalent yet underexplored form of algorithmic decision support (ADS) in this context. We conducted experiments and follow-up interviews with 60 participants, including community members, technical experts, and law enforcement agents (LEAs), to explore how lived experiences, technical knowledge, and domain expertise shape interactions with the ADS, impacting human-AI decision-making. Surprisingly, we found that domain experts (LEAs) often exhibited anchoring bias, readily accepting and engaging with the first crime map presented to them. Conversely, community members and technical experts were more inclined to engage with the tool, adjust controls, and generate different maps. Our findings highlight that all three stakeholders were able to provide critical feedback regarding AI design and use - community members questioned the core motivation of the tool, technical experts drew attention to the elastic nature of data science practice, and LEAs suggested redesign pathways such that the tool could complement their domain expertise.

Are We Asking the Right Questions?: Designing for Community Stakeholders' Interactions with AI in Policing

TL;DR

Surprisingly, it was found that domain experts (LEAs) often exhibited anchoring bias, readily accepting and engaging with the first crime map presented to them, and community members and technical experts were more inclined to engage with the tool, adjust controls, and generate different maps.

Abstract

Research into recidivism risk prediction in the criminal legal system has garnered significant attention from HCI, critical algorithm studies, and the emerging field of human-AI decision-making. This study focuses on algorithmic crime mapping, a prevalent yet underexplored form of algorithmic decision support (ADS) in this context. We conducted experiments and follow-up interviews with 60 participants, including community members, technical experts, and law enforcement agents (LEAs), to explore how lived experiences, technical knowledge, and domain expertise shape interactions with the ADS, impacting human-AI decision-making. Surprisingly, we found that domain experts (LEAs) often exhibited anchoring bias, readily accepting and engaging with the first crime map presented to them. Conversely, community members and technical experts were more inclined to engage with the tool, adjust controls, and generate different maps. Our findings highlight that all three stakeholders were able to provide critical feedback regarding AI design and use - community members questioned the core motivation of the tool, technical experts drew attention to the elastic nature of data science practice, and LEAs suggested redesign pathways such that the tool could complement their domain expertise.
Paper Structure (34 sections, 5 figures, 2 tables)

This paper contains 34 sections, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. AI in Policing and Algorithmic Crime Mapping
  4. Algorithmic Decision-Support(ADS) Systems in the Public Sector
  5. Community Members' Participation in the AI Deliberation Process
  6. Methods
  7. Application Design
  8. How does KDE Map Crime?
  9. Participants and Data Collection
  10. Participant Grouping
  11. In-Lab Activities
  12. Analysis
  13. Results
  14. Participants' Perspectives on Algorithmic Crime-Mapping (RQ1)
  15. Community members found utility in algorithmic crime mapping but were concerned about the motivations driving the use of such technologies.
  16. ...and 19 more sections

Figures (5)

  • Figure 1: Screenshot of the Algorithmic Crime Mapping Application
  • Figure 2: Algorithmic Crime Maps With Increasing Level of Complexity from Left to Right.
  • Figure 3: Boxplots of the weighted NASA TLX Scores of participants with different backgrounds
  • Figure 4: Estimates of hotspots (left) and circles (right) seen across groups and maps compared to actual values
  • Figure 5: Interaction plot of keeping default kernel settings (left), keeping default bandwidth settings (middle), default metric settings (right) across different backgrounds. Here, red points represent non-technical/community members (Group 1); green points represent technical members (Group 2); blue points represent LEAs (Group 3).