Does It Break the Presence? Using Procedurally Generated Virtual Environments for Controlled Variation in VR Experiments to Foster Generalizability

Although procedural generation (PG) has revolutionized the way designers create virtual environments (VEs), its potential to enable controlled variation in VR user studies is rarely seen. However, PG could improve the generalisability of VR experiments and reduce the influence of biases that individual designers might introduce to VEs. Nevertheless, the algorithmic generation of VEs poses a potential risk for inducing breaks in presence (BIPs). This study investigated the impact of different shadow settings within both procedurally generated and manually designed VEs. In the PG condition a unique VE was generated for each subject ( \(n=41\) ). In the control condition only one VE was used, which was built manually based on the same rules as the PG VEs. Results show that PG did not significantly affect quantitative presence ratings or lead to more breaks in presence. However, the current state of the PG used seems to increase the probability of presence scores violating normal distribution, thus there is potential for optimization to achieve more controlled variations in VEs. Nonetheless, the current PG algorithm already offers advantages in terms of time-efficiency and experimental flexibility, when compared to handmade VEs, reducing reliance on specific environmental layouts. These findings contribute to the understanding of the impact of PG on presence in VR experiments and highlight the need for further investigation into the design and implementation of PG techniques for creating VEs that effectively elicit presence.