Police implicit bias training

linear regression
logistic regression
t-test
Does implicit bias training affect the beliefs, motivations, or actions of police officers? Analyze observational data from officers attending a training program meant to increase their knowledge about biases and teach evidence-based strategies to mitigate it.
Author

Victoria Sagasta Pereira

Published

June 8, 2026

Data files
Data year

2021

Motivation

Since their creation, U.S. police departments have perpetuated unequal treatment towards black and brown individuals in their community. While overt racism is one possible explanation, researchers suspect that in ordinary day-to-day interactions between police officers and the community, implicit bias is the more likely problem.

“Implicit bias” refers to “biases that operate outside of conscious awareness and control but nevertheless influence our behaviors” (Spencer, Charbonneau, and Glaser 2016). For example, when an officer is trying to determine someone’s intentions—such as to determine if they are armed or intend to attack the officer—the officer’s judgment may follow racial stereotypes, making them more likely to draw negative conclusions about Black people. In a fast-moving situation, this may lead an officer to use force when they otherwise wouldn’t.

In an attempt to address the dangerous and often deadly consequences of implicit in policing, some departments have started integrating implicit bias training into their curriculum. However, little research has evaluated whether the trainings are effective at changing officers’ beliefs, motivations, and most importantly, actions.

This study tested a day-long “Managing Bias” diversity training program, designed by the Anti-Defamation League and delivered to sworn officers of unnamed police departments within the United States:

Throughout the workshop, officers participated in activities that raised their awareness of bias. These activities educated officers about the definition and origins of bias, the influence of culture and identity on social experience, and gaps in understanding between police officers and community members. By the end of this section, officers were expected to understand the differences between implicit and explicit bias, identify how their worldview has been shaped by their culture and identity, and appreciate how biases may affect behavior.

The final section of the training program focused on building skills to manage bias in policing. Officers were taught five practices that had been shown to reduce biased behavior. (Lai and Lisnek 2023)

Officers completed up to three surveys that assessed self-reported beliefs, attitudes, motivations, and behavior. The first survey was a baseline survey administered to officers before the training began, the second survey assessed immediate effectiveness and was administered immediately after the end of training, and the third survey was administered 1 month after training. The goal was to evaluate how the training improved officer knowledge about bias and whether it increased their use of the five strategies after the training, thus changing police behavior.

Data

The data collected is divided into two separate cohorts. In cohort 1, 4,659 officers were trained in person from July 2019 to March 2020, and in cohort 2, 3,103 officers were trained online from September 2020 to January 2021. From the first cohort, after exclusions, we obtained usable data from 3,417 officers who completed the first two surveys. From the second cohort, due to duplicate surveys, problems linking surveys together, and evidence of survey misbehavior, usable data is limited to 72 officers who completed all three surveys.1

There were many individual training sessions; each officer obtained only one. For each officer, their participation involved several steps:

  1. First, each officer completed a baseline survey before the training (time T1). This asked about their demographics, beliefs about bias, and behaviors related to bias.

  2. Next, they attended the day-long training workshop (or did so online). This taught them about implicit bias and taught several strategies for recognizing and avoiding it.

  3. Then they completed a post-training survey at the end of the workshop (time T2). This again asked about their beliefs and behaviors, as well as how they intended to use the strategies taught in the workshop.

  4. Finally, officers completed a third survey 30 days after the workshop (time T3). This asked how they had actually used the bias management strategies in their work in that time.

    However, this survey was only used for Cohort 2, and as the survey was later, many officers did not complete this survey. Other T3 surveys couldn’t be matched up to the responses from T1 or T2 and had to be excluded as well.

Each row represents one officer at one time point, so a Cohort 2 officer who was surveyed at all three times will have three rows.

Data preview

police-bias-cohort1-t1t2-long.csv

police-bias-cohort2-t1t2t3-long.csv

Variable descriptions

cohort1-t1t2-long.csv: Cohort 1, times 1 and 2
Variable Description
id Unique identifier variable
time Data collection time point (1 or 2)
cohort Describes whether data comes from the first or second cohort
knowledge Knowledge about Bias
theories Beliefs about the Malleability of Bias
biasworry Concern about Bias
strategy Use of Strategies to Manage Bias (Time 1; measured as use in the past 7 days)
strategy Intentions to Use Strategies to Manage Bias (Time 2; measured as intention to use)
strat Average of Perceived Feasibility, Efficacy, and Motivation to Use Strategies (Time 2)
centrality Police Identity Centrality (Time 1)
respect Expectations of Respect from Community Members (Time 1)
rank What is your rank? (Time 1; 0 = Below Sergeant, 1 = Sergeant or above)
gender_d What is your gender? (Time 1; 0 = Female, 1 = Male)
race What is your race or ethnicity? (Time 1; 1 = White or Caucasian, 2 = Black or African American, 3 = Hispanic or Latino, 4 = Asian or Pacific Islander, 5 = Multiracial or Other)
race_white White vs. Non-White (Time 1; 1 = White, 0 = Non-White)
race_black Black vs. Non-Black (Time 1; 1 = Black, 0 = Non-Black)
education What is the highest level of education that you have completed? (Time 1; 1 = some high school or less, 2 = high school graduate, 3 = some college, 4 = associate’s degree, 5 = bachelor’s degree, 6 = graduate degree)
education_ba Bachelor’s degree or higher vs. lower than Bachelor’s degree (Time 1; 1 = Bachelor’s degree or higher, 0 = Lower than Bachelor’s degree)
yearsworked How many years have you worked in the department? (Time 1)
frequency_white In your daily work with the Police Department, how often do you interact with people who are White or Caucasian? (Time 1; 1 = never, 2 = very rarely, 3 = rarely, 4 = occasionally, 5 = frequently, 6 = very frequently)
frequency_black In your daily work with the Police Department, how often do you interact with people who are Black or African American? (Time 1; 1 = never to 6 = very frequently)
frequency_asian In your daily work with the Police Department, how often do you interact with people who are Asian? (Time 1; 1 = never to 6 = very frequently)
frequency_hispanic In your daily work with the Police Department, how often do you interact with people who are Hispanic or Latino? (Time 1; 1 = never to 6 = very frequently)
frequency_gay In your daily work with the Police Department, how often do you interact with people who are Gay, Lesbian, or Bisexual? (Time 1; 1 = never to 6 = very frequently)
frequency_trans In your daily work with the Police Department, how often do you interact with people who are Transgender? (Time 1; 1 = never to 6 = very frequently)
facilduo_gender Educator pair’s gender makeup (Time 2; mm = Male-Male, ff = Female-Female, mf = Male-Female)
facilduo_race Educator pair’s racial makeup (Time 2; ww = White-White, bw = Black-White, asianduo = Has an Asian educator)
facilduo_le Has an ex-law-enforcement educator vs. doesn’t have an ex-law enforcement educator (Time 2; 1 = Has an ex-law-enforcement educator, 0 = Doesn’t)
facilduo_relatedle Has a close relationship with a member of law enforcement, including being one, vs. doesn’t have a close relationship (Time 2; 1 = Has a close relationship, 0 = Doesn’t)
facilduo_relatedle_nole Has a close relationship with a member of law enforcement vs. doesn’t, excluding educator who was ex-law-enforcement (Time 2; 1 = Has a close relationship, 0 = Doesn’t)
attendance Number of officers attending the session (Time 2)
substitution Substitution strategy use in the past 7 days (Time 1; 0 = 0 times, 1 = 1 time, 2 = 2 times, 3 = 3 times, 4 = 4 times, 5 = 5 times, 6 = 6 or more times)
substitution Intentions to use Substitution strategy in the next 7 days (Time 2)
perspective Perspective-taking strategy use in the past 7 days (Time 1; 0 = 0 times to 6 = 6 or more times)
perspective Intentions to use Perspective-taking strategy in the next 7 days (Time 2)
exposure Diversity exposure strategy use in the past 7 days (Time 1; 0 = 0 times to 6 = 6 or more times)
exposure Intentions to use Diversity exposure strategy in the next 7 days (Time 2)
individuation Individuation strategy use in the past 7 days (Time 1; 0 = 0 times to 6 = 6 or more times)
individuation Intentions to use Individuation strategy in the next 7 days (Time 2)
mindful Mindfulness strategy use in the past 7 days (Time 1; 0 = 0 times to 6 = 6 or more times)
mindful Intentions to use Mindfulness strategy in the next 7 days (Time 2)
knowledge_bias Subtle or implicit biases influence my decision making about other people. ( 1 = strongly disagree, 2 = moderately disagree, 3 = slightly disagree, 4 = neutral, 5 = slightly agree, 6 = moderately agree, 7 = strongly agree)
knowledge_race Whether I am aware of it or not, I use a person’s race or ethnicity to form an impression of the kind of person they are. ( 1 = strongly disagree to 7 = strongly agree)
knowledge_culture My cultural background influences how I perceive other people. ( 1 = strongly disagree to 7 = strongly agree)
theories1 People have a certain amount of bias toward other people and they can’t really change that. ( 1 = strongly disagree to 7 = strongly agree)
theories2 People’s level of bias toward other people is something very basic about them that they can’t change very much. ( 1 = strongly disagree to 7 = strongly agree)
biasworry_white I worry that I act in an unintentionally biased way toward White people. (Time 1–2; 1 = strongly disagree to 7 = strongly agree)
biasworry_black I worry that I act in an unintentionally biased way toward Black people. ( 1 = strongly disagree to 7 = strongly agree)
biasworry_asian I worry that I act in an unintentionally biased way toward Asian people. ( 1 = strongly disagree to 7 = strongly agree)
biasworry_hispanic I worry that I act in an unintentionally biased way toward Hispanic or Latino people. ( 1 = strongly disagree to 7 = strongly agree)
biasworry_lgb I worry that I act in an unintentionally biased way toward Gay, Lesbian, or Bisexual people. ( 1 = strongly disagree to 7 = strongly agree)
biasworry_trans I worry that I act in an unintentionally biased way toward Transgender people. ( 1 = strongly disagree to 7 = strongly agree)
strat_know Do you feel like you could execute these strategies correctly if you wanted to? (Time 2; 1 = definitely not, 2 = probably not, 3 = somewhat, 4 = probably, 5 = definitely)
strat_efficacy How effective do you think these strategies are for addressing issues related to bias? (Time 2; 1 = not at all, 2 = slightly, 3 = moderately, 4 = very, 5 = extremely)
strat_mot How motivated are you to use these strategies? (Time 2; 1 = not at all to 5 = extremely)
centrality1 In general, being a police officer is an important part of my self-image. (Time 1; 1 = strongly disagree to 7 = strongly agree)
centrality2 I have a strong attachment to law enforcement. (Time 1; 1 = strongly disagree to 7 = strongly agree)
centrality3 Being a police officer is a reflection of who I am. (Time 1; 1 = strongly disagree to 7 = strongly agree)
respect1 I expect respect when I put my uniform on. (Time 1; 1 = strongly disagree to 7 = strongly agree)
respect2 I expect citizens to obey my authority when I am on duty. (Time 1; 1 = strongly disagree to 7 = strongly agree)
respect3 When I wear my uniform, I expect citizens to obey any orders I give them. (Time 1; 1 = strongly disagree to 7 = strongly agree)
respect4 I expect that citizens will speak to me with respect. (Time 1; 1 = strongly disagree to 7 = strongly agree)
cohort2-t1t2t3-long.csv: Cohort 2, times 1 through 3
Variable Description
id Unique identifier variable
time Data collection time point (1, 2, or 3)
cohort Describes whether data comes from the first or second cohort
knowledge Knowledge about Bias
theories Beliefs about the Malleability of Bias
biasworry Concern about Bias
strategy Use of Strategies to Manage Bias (Time 1; measured as use in the past 7 days)
strategy Intentions to Use Strategies to Manage Bias (Time 2; measured as intention to use)
strategy Use of Strategies to Manage Bias (Time 3; measured as use in the past 7 days)
strat Average of Perceived Feasibility, Efficacy, and Motivation to Use Strategies (Time 2–3)
centrality Police Identity Centrality (Time 1)
respect Expectations of Respect from Community Members (Time 1)
rank What is your rank? (Time 1; 0 = Below Sergeant, 1 = Sergeant or above)
gender_d What is your gender? (Time 1; 0 = Female, 1 = Male)
race What is your race or ethnicity? (Time 1; 1 = White or Caucasian, 2 = Black or African American, 3 = Hispanic or Latino, 4 = Asian or Pacific Islander, 5 = Multiracial or Other)
race_white White vs. Non-White (Time 1; 1 = White, 0 = Non-White)
race_black Black vs. Non-Black (Time 1; 1 = Black, 0 = Non-Black)
education What is the highest level of education that you have completed? (Time 1; 1 = some high school or less, 2 = high school graduate, 3 = some college, 4 = associate’s degree, 5 = bachelor’s degree, 6 = graduate degree)
education_ba Bachelor’s degree or higher vs. lower than Bachelor’s degree (Time 1; 1 = Bachelor’s degree or higher, 0 = Lower than Bachelor’s degree)
yearsworked How many years have you worked in the department? (Time 1)
frequency_white In your daily work with the Police Department, how often do you interact with people who are White or Caucasian? (Time 1; 1 = never, 2 = very rarely, 3 = rarely, 4 = occasionally, 5 = frequently, 6 = very frequently)
frequency_black In your daily work with the Police Department, how often do you interact with people who are Black or African American? (Time 1; 1 = never to 6 = very frequently)
frequency_asian In your daily work with the Police Department, how often do you interact with people who are Asian? (Time 1; 1 = never to 6 = very frequently)
frequency_hispanic In your daily work with the Police Department, how often do you interact with people who are Hispanic or Latino? (Time 1; 1 = never to 6 = very frequently)
frequency_gay In your daily work with the Police Department, how often do you interact with people who are Gay, Lesbian, or Bisexual? (Time 1; 1 = never to 6 = very frequently)
frequency_trans In your daily work with the Police Department, how often do you interact with people who are Transgender? (Time 1; 1 = never to 6 = very frequently)
facilduo_gender Educator pair’s gender makeup (Time 2; mm = Male-Male, ff = Female-Female, mf = Male-Female)
facilduo_race Educator pair’s racial makeup (Time 2; ww = White-White, bw = Black-White, asianduo = Has an Asian educator)
facilduo_le Has an ex-law-enforcement educator vs. doesn’t have an ex-law enforcement educator (Time 2; 1 = Has an ex-law-enforcement educator, 0 = Doesn’t)
facilduo_relatedle Has a close relationship with a member of law enforcement, including being one, vs. doesn’t have a close relationship (Time 2; 1 = Has a close relationship, 0 = Doesn’t)
facilduo_relatedle_nole Has a close relationship with a member of law enforcement vs. doesn’t, excluding educator who was ex-law-enforcement (Time 2; 1 = Has a close relationship, 0 = Doesn’t)
attendance Number of officers attending the session (Time 2)
substitution Substitution strategy use in the past 7 days (Time 1; 0 = 0 times, 1 = 1 time, 2 = 2 times, 3 = 3 times, 4 = 4 times, 5 = 5 times, 6 = 6 or more times)
substitution Intentions to use Substitution strategy in the next 7 days (Time 2)
substitution Substitution strategy use in the past 7 days (Time 3; 0 = 0 times to 6 = 6 or more times)
perspective Perspective-taking strategy use in the past 7 days (Time 1; 0 = 0 times to 6 = 6 or more times)
perspective Intentions to use Perspective-taking strategy in the next 7 days (Time 2)
perspective Perspective-taking strategy use in the past 7 days (Time 3; 0 = 0 times to 6 = 6 or more times)
exposure Diversity exposure strategy use in the past 7 days (Time 1; 0 = 0 times to 6 = 6 or more times)
exposure Intentions to use Diversity exposure strategy in the next 7 days (Time 2)
exposure Diversity exposure strategy use in the past 7 days (Time 3; 0 = 0 times to 6 = 6 or more times)
individuation Individuation strategy use in the past 7 days (Time 1; 0 = 0 times to 6 = 6 or more times)
individuation Intentions to use Individuation strategy in the next 7 days (Time 2)
individuation Individuation strategy use in the past 7 days (Time 3; 0 = 0 times to 6 = 6 or more times)
mindful Mindfulness strategy use in the past 7 days (Time 1; 0 = 0 times to 6 = 6 or more times)
mindful Intentions to use Mindfulness strategy in the next 7 days (Time 2)
mindful Mindfulness strategy use in the past 7 days (Time 3; 0 = 0 times to 6 = 6 or more times)
knowledge_bias Subtle or implicit biases influence my decision making about other people. ( 1 = strongly disagree, 2 = moderately disagree, 3 = slightly disagree, 4 = neutral, 5 = slightly agree, 6 = moderately agree, 7 = strongly agree)
knowledge_race Whether I am aware of it or not, I use a person’s race or ethnicity to form an impression of the kind of person they are. ( 1 = strongly disagree to 7 = strongly agree)
knowledge_culture My cultural background influences how I perceive other people. ( 1 = strongly disagree to 7 = strongly agree)
theories1 People have a certain amount of bias toward other people and they can’t really change that. ( 1 = strongly disagree to 7 = strongly agree)
theories2 People’s level of bias toward other people is something very basic about them that they can’t change very much. ( 1 = strongly disagree to 7 = strongly agree)
biasworry_white I worry that I act in an unintentionally biased way toward White people. (Time 1–2; 1 = strongly disagree to 7 = strongly agree)
biasworry_black I worry that I act in an unintentionally biased way toward Black people. ( 1 = strongly disagree to 7 = strongly agree)
biasworry_asian I worry that I act in an unintentionally biased way toward Asian people. ( 1 = strongly disagree to 7 = strongly agree)
biasworry_hispanic I worry that I act in an unintentionally biased way toward Hispanic or Latino people. ( 1 = strongly disagree to 7 = strongly agree)
biasworry_lgb I worry that I act in an unintentionally biased way toward Gay, Lesbian, or Bisexual people. ( 1 = strongly disagree to 7 = strongly agree)
biasworry_trans I worry that I act in an unintentionally biased way toward Transgender people. ( 1 = strongly disagree to 7 = strongly agree)
strat_know Do you feel like you could execute these strategies correctly if you wanted to? (Time 2–3; 1 = definitely not, 2 = probably not, 3 = somewhat, 4 = probably, 5 = definitely)
strat_efficacy How effective do you think these strategies are for addressing issues related to bias? (Time 2–3; 1 = not at all, 2 = slightly, 3 = moderately, 4 = very, 5 = extremely)
strat_mot How motivated are you to use these strategies? (Time 2–3; 1 = not at all to 5 = extremely)
centrality1 In general, being a police officer is an important part of my self-image. (Time 1; 1 = strongly disagree to 7 = strongly agree)
centrality2 I have a strong attachment to law enforcement. (Time 1; 1 = strongly disagree to 7 = strongly agree)
centrality3 Being a police officer is a reflection of who I am. (Time 1; 1 = strongly disagree to 7 = strongly agree)
respect1 I expect respect when I put my uniform on. (Time 1; 1 = strongly disagree to 7 = strongly agree)
respect2 I expect citizens to obey my authority when I am on duty. (Time 1; 1 = strongly disagree to 7 = strongly agree)
respect3 When I wear my uniform, I expect citizens to obey any orders I give them. (Time 1; 1 = strongly disagree to 7 = strongly agree)
respect4 I expect that citizens will speak to me with respect. (Time 1; 1 = strongly disagree to 7 = strongly agree)

Questions

  1. Call an officer “training-resistant” if their knowledge or attitudes about implicit bias do not change from before the course to immediately afterward. Using the demographic data collected (including age, experience, or attitudes about policing), what types of officers are most likely to be training resistant?
  2. Some of the training sessions were led by an educator who had previously been a police officer. (See facilduo_le.) These educators may be better at convincing current officers of the importance of implicit bias. Do these sessions have fewer training-resistant officers? Are the training-resistant officers in those sessions different somehow, such as if these educators are better at reaching officers with particular attitudes about policing?
  3. Using each officer’s demographic data and their responses to questions about their attitudes toward policing (centrality and respect), can you predict which officers are least likely to use bias mitigation strategies before training?
  4. Does training improve actual behaviors and not just beliefs? Is there evidence the training improves the rate of bias-mitigating strategy use compared to before the training?

References

Lai, C. K., & Lisnek, J. A. (2023). “The Impact of Implicit-Bias-Oriented Diversity Training on Police Officers’ Beliefs, Motivations, and Actions.” Psychological Science, 34(4), 424-434. https://doi.org/10.1177/09567976221150617.

Data available from https://osf.io/vfdrt/overview

For background, see Spencer, Charbonneau, and Glaser (2016). “Implicit Bias and Policing.” Social and Personality Psychology Compass, 10(1), 50-63. https://doi.org/10.1111/spc3.12210

Footnotes

  1. The original study reported only 53 participants in Cohort 2 who completed all three surveys and whose data was usable. However, on matching up the data files provided, we were able to match 72 officers across all three surveys.↩︎