Validating a novel tool for coding body worn camera footage of police-community member interactions

Abstract

Objectives

This study aimed to validate a tool for coding police body-worn camera (BWC) footage to measure the dynamics of police–community encounters, including items related to community members, officer performance, and situational outcomes.

Methods

Over 1000 BWC videos were scored by five coders who participated in an iterative interrater reliability exercise to improve the agreement. Krippendorf's alphas and a multilevel simultaneous component analysis were used to assess interrater reliability and the component structure of ratings, respectively. Bootstrapped and multilevel exploratory and confirmatory factor analyses were conducted to ascertain any underlying dimensions supporting officer and community member behavior, and cluster analyses were performed to examine whether interactions could be grouped into coherent categories.

Results

The lower bound of unweighted means of bootstrapped alphas ranged from 0.67 to 0.96 depending upon item type, and 21.60% of the total variance in raters was between-subjects. Factor analyses revealed only a single factor (“defensiveness”) underlying four items related to hostility directed toward officers by community members. Cluster analyses described seven categories that were highly overlapping but conceptually plausible with cluster silhouette means ranging from −0.06 to 0.13.

Conclusions

We validated a collection of interval-level coding metrics on a random sample of over 1000 interactions between police officers and community members recorded through BWCs. In doing so, we provide a roadmap for researchers and practitioners to effectively measure the dynamics of police–community encounters and officer performance within those encounters and a framework for addressing the validity and reliability of items in future studies involving the measurement of BWC footage.