The Mixed Effects of Online Diversity Training

One-off online diversity and anti-bias trainings may be effective in changing employee attitudes, but do not change behaviors in the workplace. 


Online diversity and anti-bias trainings have become a commonly prescribed solution in response to discrimination against marginalized groups. In 2018, two Black men were subject to racial bias at a Philadelphia Starbucks when they were arrested while waiting for a friend because an employee called the police.  In response to this incident, Starbucks closed all of their US stores for one afternoon in order for employees to complete an unconscious bias training. These trainings are not uncommon in companies, as more than half of midsized and large US employers offer some form of diversity training.

Unfortunately, there are relatively few field experiments conducted to measure whether bias trainings are effective in curbing bias against women and people of color. The preliminary published meta-analyses suggest that such trainings may be more effective towards changing employee attitudes, rather than employee behaviors. Furthermore, the preliminary research shows that these trainings may have significantly different effects for different populations (i.e. for those who were originally more supportive of women vs. those who are less supportive of women).

In order to assess the effectiveness of bias trainings on employee attitudes and behaviors, researchers tested the effectiveness of a one-hour online bias training on employees of a large international organization. The study aimed to understand how to effectively reduce bias against women and people of color in employees’ attitudes and behaviors.


Online diversity trainings yielded a significant positive effect on employees’ attitudes towards women, however they did not find strong evidence of behavioral change as a result of the diversity training.

  • Although both US and non-US based participants had higher than average baseline levels of attitudinal support for women, the diversity training led to a stronger increase in non-US participants’ supportive attitudes for women and had no effect on US participants’ attitudes.
  • All employees who took the diversity training were more willing to acknowledge that their own gender biases matched those of the general population.
  • Non-US based participants who took the diversity training experienced an increase in gender-inclusive intentions by 0.206 points and there was no effect found in US based participants.
  • Three weeks after the study recruitment phase had ended, there was no significant difference in employees’ willingness to nominate women for coffee chats between employees who received the diversity training  compared to those in the control group (p = 0.914). Female US employees showed the greatest willingness to take this action (p < 0.004).
  • Six weeks after the study recruitment phase had ended, there was no significant difference in the employees’ willingness to nominate women for recognition of their excellence between the intervention and control group (p = 0.687).
  • Fourteen weeks after the study recruitment phase had ended, there was no significant overall difference for intervention group employees in their willingness to have a 15-minute meeting with male compared to female new hires (p = 0.234). However, female employees were found to be more willing to speak to female new hires (p = 0.047).

Online diversity trainings yielded the largest attitude change in those who were previously least supportive of women, but yielded the largest behavioral change in those previously most supportive of women (that is, female US employees). One-off trainings may not be an effective strategy for addressing workplace inequality due to their limited efficacy for behavior change for individuals who need it the most, those with attitudes most biased towards women.


This study conducted a field experiment at a large global organization to assess whether an online diversity training would be effective in changing employee attitudes and behaviors. In early 2017, the organization emailed 10,983 of their salaried employees worldwide to invite them to participate in a new inclusive leadership workplace training. The training was made voluntary following previous correlational research on diversity programs that showed greatest effectiveness in voluntary programs. The study tested three different kinds of trainings: gender-bias training (which gave the employee strategies to  reduce bias in their gender attitudes and behaviors), a general-bias training (which gave the employee strategies to reduce bias in their attitudes and behaviors regarding a range of social categories, including gender, race, and sexual orientation) and a placebo control training (which did not mention stereotypes and focused on psychological safety and active listening). Because the researchers found no statistically significant difference between the effectiveness of the gender-bias and the effectiveness of the general-bias training, they chose to discuss the findings from both of those trainings collectively as the results of the “diversity training.”

After a six-week recruitment period, 3,016 employees volunteered to participate in the training. Among them, 61.5% were male, 38.5% were located in the United States, and 63 countries were represented. They were each assigned to one of three trainings: gender-bias training, general-bias training, and the placebo control training. The median completion time of the training was 68 minutes. In the gender-bias and general bias trainings, employees learned about stereotypes in the workplace, received feedback on their Implicit Association Test results, and learned strategies for overcoming stereotypes. In the placebo control training, there was no mention of stereotypes, and employees learned about psychological safety and active listening. In the 20-week post-intervention follow-up, researchers assessed participant behavioral change in their willingness to nominate women for mentorship (after 3 weeks), to nominate women for recognition (after 6 weeks), and to volunteer to mentor women (after 14 weeks).

Related GAP Studies