VR and AR simulations help users practice stress management in realistic scenarios

Everyday situations can sometimes feel like big stressors, whether it’s delivering an important work presentation, attending a party full of strangers or confronting a partner. Talking to a friend or a therapist can help. But so can practice.

A new project from Carnegie Mellon University researchers aims to make that practice easier by studying whether virtual and augmented reality (VR/AR) can simulate stressful situations and help people practice stress-relief strategies. It’s a modern spin on exposure therapy: users can put on a pair of VR/AR glasses and practice what they want to say with a digital audience.

The research team, led by Anna Fang, a graduate student in the School of Computer Science’s Human-Computer Interaction Institute, will present their work at the upcoming Association for Computing Machinery Conference on Human Factors in Computing Systems (CHI 2025).

The researchers tested their stress simulation technology on a group of 19 participants, the majority of whom overwhelmingly supported it.

“For the past 10 to 20 years, virtual reality and augmented reality have taken a really big hold on the health and mental health space,” said Fang, noting the many meditation apps available for download.

However, Fang noticed that these apps usually place users in a sanitized, isolated environment—like a virtual forest or beach—while they offer tips and breathing exercises for calming down, which makes it hard to transfer those skills into the real world.

“The project comes from me wanting a practical way for people to learn these skills and apply them to their real lives,” Fang said. “Can we use virtual and augmented reality to simulate an office environment, or a conflict with someone? Then you can actually practice some of those self-care skills in an environment similar to real life.”

To start, Fang and her team focused on three scenarios that seem to cause people the most stress and anxiety in their daily lives, according to research: public speaking, crowded social events and interpersonal conflict.

The team built eight prototypes with different design elements for each of the three scenarios, for a total of 24 prototypes. Those prototypes took various forms—virtual reality, mixed or augmented reality, or a text-based environment without visual cues—and offered different degrees of interaction. The virtual audience could either respond or ask questions in one prototype, for example, or sit silently in another. In the interactive prototypes, the avatars in VR or AR could chat with users using dialog powered by a large language model. Users could also hit a button on the controller to access breathing and meditation exercises if they needed them, and instructions appeared as a popup superimposed on the scene.

“For each person, we wanted to try different designs and different combinations, so users could tell us what they liked and didn’t like,” Fang said. “The participants generally said that it was pretty realistic.”

Overall, Fang and her team found that people liked using the technology to gain more awareness about themselves. “They wanted technology that would help them learn self-sufficiency skills that they didn’t feel they had,” she said.

Participants also noted that they preferred being able to decide when the large language model suggested guidance instead of receiving that guidance automatically. They also wanted to take the headsets to other locations to help them become immersed in and more comfortable with the stressful setting. For example, a participant wanted to use the augmented headset in their home because that was where they planned to discuss an issue with their partner. Or, for public speaking, they wanted to go to the classroom the day before to practice in front of an avatar audience.

“We’re further developing it right now. We’re creating a full-fidelity deployable model that we can put on the App Store and that people can use at home,” Fang said.

In this next version, the team plans to upgrade the avatars to look more realistic and to include more text-to-speech features so the avatars can speak to users more naturally.

“If you think about being stressed in a situation, someone’s tone matters a lot,” Fang said. “We’re also giving the avatars more realistic facial expressions and movements. So if they’re angry at you, we can have them furrow their brow.”

The upcoming version will also offer an expanded menu of self-care strategies. While the current version contains mostly deep-breathing tips, the next iteration will implement relaxation and body-scanning techniques as well as grounding practices—like having people name objects around them—to help manage anxiety or panic attacks.

“We want to use the system not only to help people learn these skills, but also to experiment with different self-care strategies,” Fang said. “They can experiment in a virtual environment that works best and feels best for them, depending on the context, and then make an informed choice on what to implement in the real world.”