Classifying Harms in Social VR: A Systematic Literature Review and Taxonomy

Social VR's focus on embodied and immersive experiences has led to its increasing popularity and value for education, work, entertainment, and even our everyday social lives. However, there is also a growing concern about emerging harmful behaviours among users in social VR spaces. Despite this, existing work has highlighted a fundamental challenge in effectively identifying and mitigating such behaviours: there is no established terminology or consensus on what should be categorised as harmful behaviours in social VR. In this work, we aim to address this critical gap by providing a foundation for classifying harmful behaviours in social VR, which can help researchers, developers, and policymakers to achieve conceptual clarity and actionable insights for designing safer VR environments. Using a systematic literature review of existing work in this space, we present a comprehensive taxonomy of harmful behaviours identified in existing HCI research and classify them based on dimensions of behaviour modality (i.e., verbal, physical, environmental, and three novel system-, gesture-, and spatial-based categories) and interaction dynamics (i.e., one-to-one, one-to-many). This work provides an important conceptual contribution that enables researchers to better differentiate and systematically investigate these different types of harms in social VR. It also offers valuable guidance for the design of safer social VR environments, e.g., through informing the development of validated scales, codes of conduct, detection algorithms, and other types of interventions.