Our Work / Research & Evidence

VR DEI training at Google

Two animated avatars in business attire in conversation inside one of the VR scenarios built for Google's DEI training
A scene from one of the four VR scenarios described in the paper below.

About the paper

Equal Reality's stated role, from the paper's acknowledgements, was to help Google's team "design and build the virtual reality environments" used in the training.

Study design flow: design, sample, measure, result Four connected stages showing how the Google CHI 2024 case study was run, from designing the scenarios through to the reported results. DESIGN 4 VR scenarios Built 2021 to 2023 by 13 Google DEI volunteers with Equal Reality. SAMPLE 20 testers, 26 trainees QA cohort, then Google security team staff who opted in. 23 used Quest 2, 3 used desktop. MEASURE Optional surveys Anonymous, self-reported, after QA sessions and after training. No control or comparison group. RESULT 91% felt real, 94% satisfaction Preliminary, self-reported findings from a small sample.
How the case study ran, from scenario design through to the reported results.

What was built

The training uses four original VR scenarios, developed between 2021 and 2023 by 13 Google employee volunteers from Google's DEI programme working with a VR studio (credited elsewhere in the paper as Equal Reality). Each scenario lets a trainee experience the same difficult workplace situation from two different characters' points of view:

  • Making assumptions - a video call where a colleague's technical expertise is wrongly assumed based on appearance.
  • Everyday Diversity - two overheard office conversations involving tokenism and dismissiveness of cultural and religious needs.
  • Performance and Feedback - a performance conversation shaped by bias and stereotyping between a manager and a report.
  • In Group / Out Group - a new team member's first meeting, and assumptions about a shared cultural frame.
First-person view of hands wearing a VR headset in a generic virtual reality training scenario
Illustrative first-person VR photography. Not a frame from the Google training scenarios described here.

In each scenario, trainees can choose to "call it out," "call it in," or "ignore" the situation, then discuss the choice with a facilitator afterwards. The paper states a live, facilitated group debrief after each scenario, rather than a scripted in-app discussion, was a deliberate design choice.

Methodology

Development followed a QA phase (7 testing sessions, 20 volunteer testers, over 100 reported issues) before a first cohort of Google employees used the training. In September 2023, 32 members of Google's security team senior staff were invited to an optional session; 26 accepted, of whom 23 used Meta Quest 2 headsets and 3 used a desktop version. Feedback was collected through optional, anonymous surveys, both after the QA sessions (11 of 20 testers responded, 55%) and after the training sessions (7 of 26 trainees responded, 30%).

What the paper reports

QA testing feedback (11 of 20 testers responded)

91%

of the 11 QA testers who responded (10 of 11) said the scenarios felt real to them; all but one found the VR format effective at creating an empathic response.

82%

of the same 11 QA testers (9 of 11) rated the post-VR group discussion useful or very useful.

Training cohort feedback (7 of 26 trainees returned the survey)

94%

average satisfaction (4.7 out of 5) among all 7 trainees who returned the optional post-training survey.

82%

agreement (4.1 out of 5) that the training was useful for understanding DEI issues, from 7 of 8 responses to that item.

82%

agreement (4.1 out of 5) that the training was better than Google's standard DEI training; all but one trainee agreed.

Participants quoted in the paper described the change of viewpoint as making "microaggressions felt more real" and said the training helped them experience a situation "from the perspective of a member of a historically underrepresented group."

Limitations the authors note

Read with caution

The authors are explicit that these are preliminary, self-reported findings from a single company and a small, optional-response sample, with no comparison or control group. They list five open concerns for VR DEI training generally: hardware discomfort (weight, fit with glasses), the cost of scaling headsets and facilitators, limited visual fidelity of avatars and environments, participants wanting more agency in how scenarios unfold, and the possibility that some of the positive response reflects novelty rather than a lasting effect of the format itself.

Cited work

The paper draws on and cites earlier VR empathy and DEI-training research, including Herrera et al. (2018) on long-term empathy effects of VR perspective-taking, and Mason (2022) on VR-based DEI training. Full citations are in the published paper.

Published paper

Leveraging Virtual Reality to Enhance Diversity and Inclusion training at Google

Elie Bursztein, Karla Brown and Patrick Gage Kelley (Google) with Leonie Sanderson (Equal Reality, Australia)

CHI EA '24, ACM CHI Conference on Human Factors in Computing Systems, Honolulu, May 2024

Case study; findings described by the authors as preliminary. DOI: 10.1145/3613905.3637128

Download the paper (PDF)

Read the full paper (PDF)

Google Research publication page

CHI 2024 conference program listing

Participant sample

20 volunteer testers, QA phase

32 invited, 26 accepted, training cohort

23 Meta Quest 2, 3 desktop

Survey response rates: 55% (QA), 30% (training)