Illuminating the Unseen: A Framework for Designing and Mitigating Context-induced Harms in Behavioral Sensing

With the advanced ability to capture longitudinal sensed data and model human behavior, behavioral sensing technologies are progressing toward numerous wellbeing applications. However, the widespread use of top-down design approaches, often based on assumptions made by technology builders about user goals, needs, and preferences, can result in a lack of context sensitivity. Such oversights may lead to technologies that do not fully support the diverse needs of users and may even introduce potential harms. In this paper, we highlight two primary areas of potential harm in behavioral sensing technologies: identity-based and situation-based harms. By adopting a theory-driven approach, we propose a framework for identifying and mitigating these harms. To validate this framework, we applied it to two real-world studies of behavioral sensing as tools for systematic evaluation. Our analysis provides empirical evidence of potential harms and demonstrates the framework's effectiveness in identifying and addressing these issues. The insights derived from our evaluations, coupled with the reflection on the framework, contribute both conceptually and practically to the field. Our goal is to guide technology builders in designing more context-sensitive sensing technologies, thereby supporting responsible decision-making in this rapidly evolving field.