U.S. flag

An official website of the United States government, Department of Justice.

An Evaluation of Simulation vs. Classroom-Based Implicit Bias Training to Improve Police Decision Making and Enhance the Outcomes of Police-Citizen Encounters

NCJ Number
Date Published
21 pages

This document reports on a study aimed at evaluating the impact of implicit bias training on four distinct outcomes: officer behavior, based on body-worn camera footage; specific discrimination-based community member complaints against police; officer perceptions of training effectiveness; and arrestee perceptions of treatment by police.


This report discusses an experimental research design study that evaluated two types of training: classroom-based implicit bias training, and simulation-based counter bias training. The researchers created four separate groups of patrol officers, totaling about 400, who were followed over an approximate three-year period: first, a group that received only classroom-based training; second, a group that received only simulation-based training; third, a combined treatment group that received both types of training; and fourth, a group that received neither type of training, functioning as the control group. The four main research questions for the study were: if implicit bias training can impact police behavior by promoting fairness in their decision-making during interactions with community members; if implicit bias training can impact community member perceptions of police bias based on discrimination complaints; if classroom-based training, simulation-based training, or a combination of both is more effective at improving fairness in police decision making and reducing perceptions of bias in community members; and what officer perceptions are of implicit bias training. The author reports that all major activities, excluding the gathering of arrestee perceptions of police treatment, were accomplished; and states that analysis of performance scores revealed a significant interaction between the intervention group and time, pre- and post-scores. The author also reports that performance scores in the combined group went from an average of 50 percent before training, to 61 percent after the training. Other study procedures, results, limitations, and artifacts are also discussed.

Date Published: January 1, 2022