Detecting Social Contexts from Mobile Sensing Indicators in Virtual Interactions

Explore a 15-minute conference talk from UbiComp/ISWC 2023 that delves into detecting social contexts using mobile sensing indicators during virtual interactions with socially anxious individuals. Learn about a study conducted with highly socially anxious undergraduate students, examining the feasibility of using wristband sensors to detect experimentally manipulated virtual social contexts. Discover how a multitask machine learning pipeline leverages passively sensed biobehavioral streams to identify contexts relevant to social anxiety, including the presence of social situations, group size, degree of social evaluation, and phases of social interactions. Gain insights into the predictive accuracy of detecting various social contexts and the differential importance of sensing streams. Understand the implications for future work on context detection, including considerations for optimal sensing duration, utility of different sensing modalities, and the need for personalization in developing just-in-time adaptive interventions for social anxiety research.