My role
Team Leader (Product manager and Research Lead)
Main tools
Unity, LeanTime, Figjam, Microsoft Office Suite
Responsabilities
Project & Product Management, UX Design in VR, Academic research design
Time
Jul 2023 - Present
The What?
Designed and developed an immersive VR training programme to teach bystander intervention strategies for preventing sexual harassment. The project involved creating a realistic bar scene in Unity, utilising motion capture for authentic interactions, and conducting Jul 2023 - Present a controlled study to compare the effectiveness of VR training against traditional seminars.
The Why?
Traditional bystander training methods, such as seminars and videos, often fail to engage participants or drive significant behavioural change. Research indicates that VR enhances learning by increasing presence, agency, and emotional engagement. This project aimed to leverage VR's strengths to create a more impactful training experience while addressing the growing need for innovative, scalable intervention solutions.
The How?
Traditional bystander training methods, such as seminars and videos, often fail to engage participants or drive significant behavioural change. Research indicates that VR enhances learning by increasing presence, agency, and emotional engagement. This project aimed to leverage VR's strengths to create a more impactful training experience while addressing the growing need for innovative, scalable intervention solutions.
A 2021 survey by UN Women found that 71% of women have experienced sexual harassment in public spaces like hospitality venues, public transport, events, and online platforms. Another 8% of women didn’t respond definitively, suggesting the true prevalence could be even higher. Despite these alarming numbers, studies show that 77% of witnesses to harassment fail to intervene, highlighting a significant gap between observation and action.
This behaviour is linked to the bystander effect, where individuals are less likely to step in and help when others are present. Originally studied in the 1960s, this phenomenon shows how inaction can allow harmful situations to escalate. However, research has also demonstrated that bystander intervention can prevent harassment and offer crucial support to victims.
Traditional training methods—such as videos, seminars, and written materials—have been used extensively to encourage bystander action. A review of multiple studies has shown these approaches can effectively improve attitudes and behaviours toward preventing harassment.
Virtual Reality (VR) has emerged as a powerful tool for training in various fields, such as medicine, where it has been shown to enhance knowledge retention and practical skills. VR’s unique ability to create a sense of "Presence" (feeling physically present in a virtual environment) and "Agency" (control over one’s actions) makes learning more engaging and effective, as highlighted by the Cognitive Affective Model of Immersive Learning. These immersive qualities help learners retain training material by fostering emotional connection and active participation.
Research has shown that VR bystander training can improve participants' intentions to intervene in harassment situations, outperforming traditional 2D video training. However, past studies often used basic VR setups, such as Google Cardboard, which limited the potential of the technology. Advanced VR systems offer opportunities to further explore and maximise these benefits.
Building on this, we developed a VR bystander training programme using Unity and a flagship VR headset. The programme aims to enhance intervention intentions, knowledge about sexual harassment, attitudes towards harassment, and motivation to learn (MTL).
This study seeks to demonstrate how advanced VR technology can offer more impactful training experiences, potentially transforming how bystander intervention training is delivered. We will compare the effectiveness of VR training with traditional seminar training.
We recruited students with diverse backgrounds from the University of Bath and formed three multidisciplinary teams:
VR Team: Responsible for developing the immersive VR scene used in the training experiment.
Academic Team: Handles the research, runs the experiment, and conducts data analysis.
Machine Learning (ML) Team: Utilises data collected during the experiment to train an ML model aimed at automating the training experience. This will replace the manual input required in the first iteration of the training system.
To coordinate these teams, I facilitated planning meetings to ensure alignment across disciplines, accounting for each member’s time and expertise. This collaborative stage ensured all team members shared a unified vision and clearly understood their responsibilities. Together, we developed the project’s mission, vision, and core values to guide our work and decision-making process.
Vision Statement
To revolutionise bystander intervention training through innovative VR technology, fostering a world where individuals are empowered to recognise, respond to, and prevent crimes, creating safer and more inclusive public spaces.
Mission Statement
Our mission is to design, implement, and evaluate an immersive VR-based training system that enhances participants’ knowledge, attitudes, and motivation to act as proactive bystanders. By combining cutting-edge technology with evidence-based practices, we aim to address gaps in traditional training methods and drive meaningful behavioural change.
While I served as the team leader, I recognised that others were more technically skilled in development. My primary role focused on communication, understanding team members' needs and challenges, and facilitating support where needed. I also wanted this project to be a learning opportunity for everyone involved, introducing agile practices to help the team organise their work more effectively and gain industry-relevant experience.
Kanban for Clarity and Collaboration:
We adopted a Kanban board instead of traditional to-do lists, which transformed how the team worked. Members found this approach far more collaborative and transparent, allowing everyone to see progress, priorities, and bottlenecks at a glance.
Midweek Stand-Ups for Early Problem-Solving:
In addition to weekly team meetings, we introduced a midweek stand-up. This short check-in followed agile principles, helping us identify obstacles early and address issues before the main meeting. This practice significantly improved our productivity and progress.
Iterative Development with MVPs:
Instead of attempting to develop the entire VR scene in one go, we focused on creating Minimum Viable Products (MVPs). These basic versions allowed us to test the viability of the training system early, refine the design, and solve potential problems as they arose.
These practices not only improved the efficiency and collaboration of the team but also provided valuable hands-on experience with industry-standard workflows.
Using Unity, we created a bar scene as the central environment for the training programme. Once the first Minimum Viable Product (MVP) of the bar scene was complete, we collaborated with the Centre for the Analysis of Motion, Entertainment Research, and Applications (CAMERA) team to conduct motion capture for our NPCs.
Given the specific actions required for the training programme, we chose to create custom animations through motion capture instead of relying on pre-made assets. To ensure the interactions felt authentic and realistic, we invited drama and acting students from Bath Spa University to perform the scenes during the motion capture sessions.
This approach allowed us to create a highly immersive and believable training environment, aligning with our goal of delivering an impactful VR experience.
Given the sensitive nature of sexual harassment, we recognised that our student teams might lack the expertise to ensure the appropriateness of the entire experiment. To address this, we collaborated with Somerset and Avon Rape and Sexual Abuse Support (SARSAS), an external organisation specialising in this area.
Through this partnership, SARSAS provided training and guidance to our team on handling the subject matter responsibly. Additionally, we shared all materials and content with them for review, ensuring the VR training experience was both effective and safe for participants. This collaboration was crucial in maintaining ethical standards and delivering a meaningful intervention.
To demonstrate the effectiveness of the VR bystander training, we will run a control group where participants will undergo traditional seminar-based training. We will measure key areas—such as intervention intentions, knowledge about sexual harassment, attitudes towards harassment, and motivation to learn (MTL)—both before and after the training. The results will be analysed using Pearson correlation and an independent sample t-test to determine whether there is a significant difference between the two training methods.
Based on previous research, we hypothesise that the VR training will be more effective. This is due to VR’s ability to enhance learning by improving presence (the feeling of being physically present in a virtual environment), agency (the sense of control over one’s actions), and emotional engagement, all of which are known to drive greater knowledge retention and behavioural change.
Currently, the VR scene operates with input from an experimenter. Based on the participant’s actions, the experimenter classifies their behaviour into one of the 5Ds (Distract, Delegate, Document, Direct, or Delay), and the NPCs in the scene respond accordingly.
To make this training more practical and accessible, the goal is to automate this process. The Machine Learning (ML) team is currently training a model using online datasets to classify participant behaviour and enable the NPCs to react dynamically without the need for manual input. Once the experiment is complete, we plan to train the model using higher-quality, context-specific data collected during the study, ensuring greater accuracy and relevance for this use case.
This project is currently ongoing, with the final report of results expected to be released in August 2024.
If you are in the Bath area and interested in participating, feel free to contact me at yiran.ivan.liu@gmail.com.
For more information about the science behind this project, don’t hesitate to reach out—I’d be happy to discuss it further.