Two Dire Straits concert performances were recreated to study audience responses. Surprisingly, the surrounding virtual audience, rather than the band, was the most impactful factor. Sentiment analysis of participant essays revealed that some participants, in particular women, became concerned because they thought that virtual men in the crowd around them would approach them. Others become concerned because they thought that members of the virtual crowd were ‘staring’ at them. Thus although there was high presence, there was, in some cases, low sentiment. This would never have been discovered by traditional methods focussing on questionnaires.
The project developed a novel method (Adaptive Multimodal Matching – A3M) to find optimal VR settings for presence in real-time based on participant choices, rather than post-experience questionnaires. A3M, a Reinforcement Learning (AI) model, successfully determined optimal configurations for presence. An experiment in conjunction with Facebook was used to determine optimal configurations for a virtual TED talk.
VR reminiscence therapy was explored, with older adults reliving a salient Eurovision performance from their youth. While analysis is pending, pilot results suggest these immersive experiences can positively impact negative aspects of ageing.
The project also demonstrated how VR can be a potent tool to address biases and encourage pro-social behavior. We enhanced earlier findings that embodying people in virtual bodies of a different race can reduce their implicit biases against people of that race. Here we found an increase in implicit bias when the embodiment occurred in a stressful scenario, challenging the notion of VR as an automatic empathy machine. Continuing this line, our ‘Golden Rule Embodiment Paradigm’, where participants witness their own negative actions or acquiescent inaction from the embodied perspective of a victim, was successful in reducing bias and encouraging helpful interventions in a VR scenario in a United States police department, a project in collaboration with Google Jigsaw (
https://youtu.be/tb9QAUkZWic(opens in new window)). This method is widely used by our spin-off company kiin.tech to address workplace discrimination.
Several technical advances were made: a system to create realistic virtual bodies from photos, a solution for natural walking movement in VR despite space constraints, and the QuickVR programming library. The VR United software allows remote users to interact in VR with with others with each person with a virtual body that looks like themselves. It facilitated journalistic interviews (
https://www.youtube.com/watch?v=1dACicAYdYg(opens in new window) https://youtu.be/njVlI8409fs(opens in new window)) and a conference panel discussion including Albert Einstein driven by ChatGPT (
https://youtu.be/qkN1F3QAhp8(opens in new window)).
We went beyond the original project specification by studying ‘Change blindness’, a phenomenon where people can be oblivious to major changes in their environment. We found that in VR people were surprisingly oblivious to gradual changes in the appearances of themselves and others (
https://youtu.be/XPkUIjBKqUU(opens in new window)). This concept was applied to a single-session therapy for helping people overcome their fear of public speaking, in comparison with a traditional 5 session exposure therapy.
We developed a new technique that allows participants to walk through a VR scenario even though the physical space in which they are located is much smaller than the virtual space in which they can walk.
Finally, we found that gradually increasing an environment from dark to light allows faster movement through the environment, but without causing simulator sickness. This is very important because it offers a way forward for people to use VR without feeling the symptoms of simulator sickness.
