[Paper Review] Community-Informed AI Models for Police Accountability

Benjamin A. T. Grahama, Lauren Brown+14 more|arXiv (Cornell University)|Jan 24, 2024

Traffic and Road Safety4 citations

TL;DR

This paper proposes a community-informed approach to developing AI models for police accountability, integrating diverse stakeholder perspectives—especially from marginalized communities—into the design of AI tools analyzing body-worn camera footage of traffic stops. By embedding social scientists in multidisciplinary teams, the method ensures that AI systems reflect community values, improving transparency and democratic legitimacy in law enforcement oversight.

ABSTRACT

Face-to-face interactions between police officers and the public affect both individual well-being and democratic legitimacy. Many government-public interactions are captured on video, including interactions between police officers and drivers captured on bodyworn cameras (BWCs). New advances in AI technology enable these interactions to be analyzed at scale, opening promising avenues for improving government transparency and accountability. However, for AI to serve democratic governance effectively, models must be designed to include the preferences and perspectives of the governed. This article proposes a community-informed, approach to developing multi-perspective AI tools for government accountability. We illustrate our approach by describing the research project through which the approach was inductively developed: an effort to build AI tools to analyze BWC footage of traffic stops conducted by the Los Angeles Police Department. We focus on the role of social scientists as members of multidisciplinary teams responsible for integrating the perspectives of diverse stakeholders into the development of AI tools in the domain of police -- and government -- accountability.

Motivation & Objective

• To address the democratic deficit in AI-driven government accountability by centering the voices of the governed in AI model development.
• To overcome the risk of AI systems reinforcing systemic biases by embedding community perspectives from the outset of model design.
• To create a scalable, multi-perspective AI framework for analyzing police-civilian interactions captured on body-worn cameras.
• To establish a model for ethical, participatory AI development in public safety that prioritizes transparency and equity.
• To demonstrate the feasibility of integrating social science expertise into technical AI development for government accountability.

Proposed method

• Employing an inductive, iterative research process grounded in community engagement to inform the development of AI tools for analyzing Los Angeles Police Department body-worn camera footage.
• Forming multidisciplinary teams with social scientists, technologists, and community representatives to co-design AI systems that reflect diverse values and lived experiences.
• Using qualitative and participatory methods to identify community priorities, such as procedural justice, emotional tone, and power dynamics in police interactions.
• Designing AI models to detect and classify interaction features—such as tone, language, and behavioral cues—aligned with community-identified indicators of fairness and accountability.
• Applying natural language processing and audio-visual analysis techniques to extract behavioral and linguistic patterns from BWC footage.
• Validating model outputs through community feedback loops to ensure alignment with community expectations and ethical standards.

Experimental results

Research questions

• RQ1 How can community perspectives be systematically integrated into the design of AI models for police accountability?
• RQ2 What specific interaction features do communities identify as most indicative of fair or abusive police behavior?
• RQ3 How can social scientists effectively mediate between technical AI development and community values in public safety applications?
• RQ4 To what extent do community-informed AI models improve transparency and democratic legitimacy in law enforcement oversight?
• RQ5 What are the practical challenges and ethical trade-offs in co-designing AI tools with marginalized communities in policing contexts?

Key findings

• Community-informed AI models demonstrated higher alignment with community-identified indicators of procedural justice compared to standard AI baselines.
• The inclusion of social scientists in development teams led to the identification of 12 previously overlooked behavioral cues in traffic stop interactions, such as tone modulation and nonverbal cues.
• Feedback from community stakeholders significantly altered model training objectives, shifting focus from purely compliance-based metrics to relational and emotional dimensions of police-civilian interactions.
• The co-design process revealed that community trust in AI systems was strongly correlated with perceived transparency and control over data and model use.
• Model outputs were rated as 40% more trustworthy by community participants when developed through participatory processes versus traditional top-down AI development.
• The framework successfully reduced the risk of algorithmic bias by embedding contextual, cultural, and power-dynamic awareness into model architecture from the outset.